diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,93016 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8445025607497003,
+  "global_step": 15500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4782608695652175e-06,
+      "loss": 11.037,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.217391304347826e-06,
+      "loss": 11.0295,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.956521739130435e-06,
+      "loss": 10.9861,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.695652173913044e-06,
+      "loss": 10.914,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0434782608695651e-05,
+      "loss": 10.884,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2173913043478261e-05,
+      "loss": 10.8493,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.391304347826087e-05,
+      "loss": 10.8226,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.565217391304348e-05,
+      "loss": 10.8693,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.739130434782609e-05,
+      "loss": 10.5155,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9130434782608697e-05,
+      "loss": 10.399,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0869565217391303e-05,
+      "loss": 10.4075,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2608695652173914e-05,
+      "loss": 10.4043,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4347826086956523e-05,
+      "loss": 10.1528,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.608695652173913e-05,
+      "loss": 10.1586,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.782608695652174e-05,
+      "loss": 9.7757,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9565217391304352e-05,
+      "loss": 9.5792,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.130434782608696e-05,
+      "loss": 10.1857,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.304347826086956e-05,
+      "loss": 10.4012,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.478260869565218e-05,
+      "loss": 9.881,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6521739130434786e-05,
+      "loss": 10.0128,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.8260869565217395e-05,
+      "loss": 10.1196,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-05,
+      "loss": 10.073,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.1739130434782605e-05,
+      "loss": 9.5795,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.347826086956522e-05,
+      "loss": 9.5711,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.521739130434783e-05,
+      "loss": 9.7006,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.695652173913044e-05,
+      "loss": 9.6825,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.8695652173913046e-05,
+      "loss": 9.6489,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.0434782608695655e-05,
+      "loss": 9.7106,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.217391304347826e-05,
+      "loss": 10.0701,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.391304347826087e-05,
+      "loss": 9.1305,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.565217391304348e-05,
+      "loss": 9.3834,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.739130434782609e-05,
+      "loss": 9.1854,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.9130434782608704e-05,
+      "loss": 8.9188,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.086956521739131e-05,
+      "loss": 8.8814,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.260869565217392e-05,
+      "loss": 9.2415,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.434782608695652e-05,
+      "loss": 8.9918,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.608695652173912e-05,
+      "loss": 8.6699,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.782608695652173e-05,
+      "loss": 8.7858,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.956521739130436e-05,
+      "loss": 9.167,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.130434782608696e-05,
+      "loss": 8.716,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.304347826086957e-05,
+      "loss": 9.2203,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.478260869565218e-05,
+      "loss": 9.1121,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.652173913043479e-05,
+      "loss": 8.8143,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.82608695652174e-05,
+      "loss": 8.4624,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-05,
+      "loss": 8.5799,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.173913043478262e-05,
+      "loss": 9.1277,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.347826086956521e-05,
+      "loss": 9.1723,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.521739130434783e-05,
+      "loss": 7.9242,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.695652173913044e-05,
+      "loss": 8.2779,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.869565217391305e-05,
+      "loss": 8.8725,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.043478260869566e-05,
+      "loss": 7.887,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.217391304347827e-05,
+      "loss": 8.4189,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.391304347826087e-05,
+      "loss": 8.118,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.565217391304348e-05,
+      "loss": 8.5339,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.739130434782609e-05,
+      "loss": 8.5874,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.91304347826087e-05,
+      "loss": 7.8053,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010086956521739131,
+      "loss": 7.9836,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010260869565217393,
+      "loss": 8.1102,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010434782608695653,
+      "loss": 7.5594,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010608695652173915,
+      "loss": 8.026,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010782608695652174,
+      "loss": 8.2103,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010956521739130434,
+      "loss": 9.1035,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011130434782608696,
+      "loss": 7.885,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011304347826086956,
+      "loss": 8.0408,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011478260869565218,
+      "loss": 8.06,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011652173913043479,
+      "loss": 7.9785,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011826086956521741,
+      "loss": 7.4277,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012,
+      "loss": 7.8381,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012173913043478263,
+      "loss": 7.2277,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012347826086956523,
+      "loss": 7.0097,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012521739130434784,
+      "loss": 7.9914,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012695652173913045,
+      "loss": 7.6968,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012869565217391303,
+      "loss": 8.6249,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013043478260869567,
+      "loss": 7.3139,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013217391304347825,
+      "loss": 7.4093,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013391304347826088,
+      "loss": 8.3534,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013565217391304347,
+      "loss": 7.7112,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001373913043478261,
+      "loss": 8.63,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001391304347826087,
+      "loss": 7.3533,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014086956521739132,
+      "loss": 6.699,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014260869565217393,
+      "loss": 7.4764,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014434782608695654,
+      "loss": 7.3639,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014608695652173914,
+      "loss": 7.5441,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014782608695652173,
+      "loss": 6.7957,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014956521739130436,
+      "loss": 7.5739,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015130434782608694,
+      "loss": 7.7414,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015304347826086958,
+      "loss": 6.7718,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001547826086956522,
+      "loss": 7.3006,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001565217391304348,
+      "loss": 8.15,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001582608695652174,
+      "loss": 7.1952,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00016,
+      "loss": 6.8761,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00016173913043478262,
+      "loss": 7.4032,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00016347826086956523,
+      "loss": 6.3697,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00016521739130434784,
+      "loss": 7.2093,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00016695652173913042,
+      "loss": 7.0734,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00016869565217391306,
+      "loss": 7.5401,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00017043478260869566,
+      "loss": 7.0077,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00017217391304347827,
+      "loss": 7.2093,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00017391304347826088,
+      "loss": 7.4396,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001756521739130435,
+      "loss": 6.9487,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001773913043478261,
+      "loss": 7.5511,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001791304347826087,
+      "loss": 9.2023,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018086956521739132,
+      "loss": 8.5675,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018260869565217392,
+      "loss": 6.8039,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018434782608695653,
+      "loss": 7.127,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018608695652173914,
+      "loss": 8.0662,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018782608695652175,
+      "loss": 7.8376,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018956521739130436,
+      "loss": 6.6642,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018956521739130436,
+      "loss": 8.1446,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019130434782608697,
+      "loss": 8.0923,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019304347826086958,
+      "loss": 7.3934,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019478260869565218,
+      "loss": 8.0433,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001965217391304348,
+      "loss": 7.3966,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001982608695652174,
+      "loss": 6.9985,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002,
+      "loss": 7.3831,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999998961619109,
+      "loss": 7.3955,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999958464785923,
+      "loss": 7.221,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999906545849197,
+      "loss": 7.1287,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999983385948873,
+      "loss": 6.8499,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999740405855478,
+      "loss": 6.9933,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999626185143518,
+      "loss": 5.8197,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999491197590066,
+      "loss": 7.3202,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999335443475453,
+      "loss": 8.5134,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999158923123144,
+      "loss": 7.5446,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019998961636899734,
+      "loss": 6.9415,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019998743585214932,
+      "loss": 7.2521,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019998504768521587,
+      "loss": 6.8303,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019998245187315662,
+      "loss": 7.5208,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999796484213624,
+      "loss": 7.3127,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019997663733565537,
+      "loss": 6.5161,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019997341862228884,
+      "loss": 7.8396,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019996999228794727,
+      "loss": 6.7202,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019996635833974637,
+      "loss": 6.9397,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000199962516785233,
+      "loss": 6.9117,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019995846763238512,
+      "loss": 7.228,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999542108896119,
+      "loss": 5.9583,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999497465657535,
+      "loss": 7.556,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019994507467008133,
+      "loss": 6.8179,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019994019521229776,
+      "loss": 7.5084,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999351082025363,
+      "loss": 6.7886,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019992981365136143,
+      "loss": 7.1628,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999243115697687,
+      "loss": 8.4492,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999186019691846,
+      "loss": 7.4621,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019991268486146662,
+      "loss": 6.2506,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019990656025890315,
+      "loss": 7.8483,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019990022817421358,
+      "loss": 6.7205,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019989368862054812,
+      "loss": 7.3419,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019988694161148787,
+      "loss": 6.9218,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019987998716104472,
+      "loss": 7.3075,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019987282528366148,
+      "loss": 8.0822,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001998654559942116,
+      "loss": 6.7086,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019985787930799939,
+      "loss": 7.3521,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019985009524075978,
+      "loss": 5.8272,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001998421038086584,
+      "loss": 6.9208,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019983390502829166,
+      "loss": 7.1775,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019982549891668639,
+      "loss": 7.1362,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019981688549130006,
+      "loss": 6.9703,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019980806477002078,
+      "loss": 6.4022,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019979903677116702,
+      "loss": 7.2608,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019978980151348782,
+      "loss": 7.2436,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001997803590161626,
+      "loss": 7.4781,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019977070929880116,
+      "loss": 7.5744,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001997608523814437,
+      "loss": 6.0182,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019975078828456065,
+      "loss": 8.1689,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019974051702905277,
+      "loss": 6.9703,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000199730038636251,
+      "loss": 7.0568,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019971935312791645,
+      "loss": 6.3952,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019970846052624043,
+      "loss": 6.3805,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001996973608538442,
+      "loss": 7.0803,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019968605413377925,
+      "loss": 6.8729,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019967454038952677,
+      "loss": 5.9451,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019966281964499825,
+      "loss": 6.9589,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019965089192453475,
+      "loss": 6.1251,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019963875725290738,
+      "loss": 6.7885,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019962641565531692,
+      "loss": 6.7775,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019961386715739395,
+      "loss": 6.3879,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001996011117851987,
+      "loss": 6.7256,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019958814956522108,
+      "loss": 5.6447,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019957498052438046,
+      "loss": 6.1073,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019956160469002584,
+      "loss": 6.2954,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019954802208993564,
+      "loss": 6.8668,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001995342327523177,
+      "loss": 6.6517,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019952023670580916,
+      "loss": 6.5772,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001995060339794765,
+      "loss": 5.8638,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001994916246028154,
+      "loss": 7.4892,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019947700860575065,
+      "loss": 6.8524,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019946218601863625,
+      "loss": 5.7346,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019944715687225518,
+      "loss": 7.3514,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019943192119781938,
+      "loss": 6.3762,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019941647902696973,
+      "loss": 7.7092,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019940083039177593,
+      "loss": 6.5555,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019938497532473646,
+      "loss": 7.0201,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001993689138587785,
+      "loss": 6.5948,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019935264602725797,
+      "loss": 6.9813,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019933617186395917,
+      "loss": 6.3278,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001993194914030951,
+      "loss": 5.4112,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019930260467930706,
+      "loss": 6.333,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019928551172766477,
+      "loss": 7.1135,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019926821258366622,
+      "loss": 6.089,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019925070728323758,
+      "loss": 6.5969,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019923299586273324,
+      "loss": 5.7653,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019921507835893553,
+      "loss": 6.6566,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001991969548090549,
+      "loss": 5.8675,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001991786252507297,
+      "loss": 5.7053,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019916008972202585,
+      "loss": 6.9698,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019914134826143747,
+      "loss": 6.9024,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019912240090788595,
+      "loss": 6.5083,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019910324770072048,
+      "loss": 6.8729,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019908388867971772,
+      "loss": 6.7727,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019906432388508175,
+      "loss": 6.4347,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019904455335744393,
+      "loss": 7.6809,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000199024577137863,
+      "loss": 7.2713,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019900439526782479,
+      "loss": 6.1903,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001989840077892422,
+      "loss": 6.5933,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019896341474445525,
+      "loss": 6.2297,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019894261617623072,
+      "loss": 6.0891,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001989216121277623,
+      "loss": 6.0859,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019890040264267035,
+      "loss": 6.2337,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019887898776500202,
+      "loss": 7.558,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019885736753923083,
+      "loss": 6.6039,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019883554201025687,
+      "loss": 5.7272,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019881351122340654,
+      "loss": 6.0737,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019879127522443255,
+      "loss": 6.6436,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019876883405951377,
+      "loss": 6.4725,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019874618777525517,
+      "loss": 6.8742,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019872333641868768,
+      "loss": 5.9682,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001987002800372681,
+      "loss": 7.6419,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019867701867887908,
+      "loss": 6.0549,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001986535523918289,
+      "loss": 7.0413,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019862988122485144,
+      "loss": 6.6702,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001986060052271061,
+      "loss": 6.1325,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001985819244481776,
+      "loss": 5.7842,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000198557638938076,
+      "loss": 5.8543,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019853314874723657,
+      "loss": 7.4494,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001985084539265195,
+      "loss": 5.9109,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019848355452721012,
+      "loss": 6.8072,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019845845060101856,
+      "loss": 5.9308,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019843314220007968,
+      "loss": 7.2163,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019840762937695295,
+      "loss": 5.225,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001983819121846225,
+      "loss": 6.8089,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019835599067649678,
+      "loss": 5.7091,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019832986490640856,
+      "loss": 8.2218,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001983035349286149,
+      "loss": 7.2755,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019827700079779685,
+      "loss": 6.9782,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019825026256905948,
+      "loss": 6.8109,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019822332029793173,
+      "loss": 6.7142,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019819617404036628,
+      "loss": 5.8663,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019816882385273945,
+      "loss": 6.5502,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019814126979185104,
+      "loss": 7.4314,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001981135119149243,
+      "loss": 6.9368,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019808555027960568,
+      "loss": 7.3373,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001980573849439649,
+      "loss": 6.9256,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001980290159664946,
+      "loss": 7.0849,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001980004434061104,
+      "loss": 7.7046,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019797166732215076,
+      "loss": 5.734,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019794268777437667,
+      "loss": 7.019,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019791350482297178,
+      "loss": 5.9855,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019788411852854213,
+      "loss": 6.3863,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019785452895211604,
+      "loss": 6.6818,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019782473615514404,
+      "loss": 5.9488,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019779474019949862,
+      "loss": 6.6048,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001977645411474743,
+      "loss": 6.5501,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001977341390617873,
+      "loss": 5.9268,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019770353400557546,
+      "loss": 6.0395,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019767272604239824,
+      "loss": 6.4898,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001976417152362364,
+      "loss": 6.5003,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019761050165149208,
+      "loss": 6.1787,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019757908535298832,
+      "loss": 8.5834,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019754746640596944,
+      "loss": 7.482,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019751564487610038,
+      "loss": 6.5532,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001974836208294669,
+      "loss": 6.1076,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019745139433257526,
+      "loss": 5.0299,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001974189654523523,
+      "loss": 6.6506,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000197386334256145,
+      "loss": 6.6508,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019735350081172067,
+      "loss": 7.506,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001973204651872665,
+      "loss": 6.0959,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001972872274513896,
+      "loss": 6.5157,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019725378767311685,
+      "loss": 5.3775,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001972201459218947,
+      "loss": 6.5278,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019718630226758907,
+      "loss": 7.4372,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019715225678048514,
+      "loss": 6.621,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001971180095312873,
+      "loss": 6.1439,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001970835605911189,
+      "loss": 5.4567,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001970489100315222,
+      "loss": 6.7354,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019701405792445814,
+      "loss": 6.2487,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001969790043423063,
+      "loss": 6.8649,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019694374935786457,
+      "loss": 7.7255,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019690829304434916,
+      "loss": 6.2485,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019687263547539437,
+      "loss": 6.4324,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019683677672505253,
+      "loss": 6.4311,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019680071686779366,
+      "loss": 5.2853,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019676445597850557,
+      "loss": 6.8684,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001967279941324934,
+      "loss": 6.5609,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019669133140547978,
+      "loss": 7.0113,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001966544678736044,
+      "loss": 6.9425,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019661740361342415,
+      "loss": 5.3716,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019658013870191258,
+      "loss": 7.2262,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019654267321646,
+      "loss": 6.4065,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019650500723487335,
+      "loss": 6.1945,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019646714083537594,
+      "loss": 7.6613,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001964290740966072,
+      "loss": 6.9204,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019639080709762268,
+      "loss": 6.0384,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001963523399178939,
+      "loss": 6.3878,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001963136726373079,
+      "loss": 7.5441,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001962748053361675,
+      "loss": 6.4678,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001962357380951908,
+      "loss": 5.9761,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019619647099551116,
+      "loss": 6.3967,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000196157004118677,
+      "loss": 7.4042,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001961173375466516,
+      "loss": 5.7492,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019607747136181298,
+      "loss": 6.2626,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019603740564695372,
+      "loss": 6.5219,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019599714048528077,
+      "loss": 6.091,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019595667596041528,
+      "loss": 6.7496,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001959160121563924,
+      "loss": 7.0367,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019587514915766124,
+      "loss": 5.5604,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019583408704908442,
+      "loss": 6.5306,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001957928259159382,
+      "loss": 7.3524,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019575136584391213,
+      "loss": 5.9081,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001957097069191089,
+      "loss": 6.2148,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019566784922804415,
+      "loss": 6.9661,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019562579285764637,
+      "loss": 7.4583,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019558353789525662,
+      "loss": 6.5136,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019554108442862837,
+      "loss": 6.096,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019549843254592736,
+      "loss": 6.0048,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019545558233573134,
+      "loss": 6.2904,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001954125338870301,
+      "loss": 6.9422,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019536928728922494,
+      "loss": 8.4093,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019532584263212878,
+      "loss": 5.0343,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019528220000596575,
+      "loss": 7.4,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019523835950137128,
+      "loss": 6.1184,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001951943212093916,
+      "loss": 6.7497,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019515008522148376,
+      "loss": 5.9455,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019510565162951537,
+      "loss": 6.4695,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001950610205257644,
+      "loss": 6.433,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019501619200291908,
+      "loss": 6.3889,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001949711661540775,
+      "loss": 6.0219,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001949259430727477,
+      "loss": 6.4548,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019488052285284715,
+      "loss": 6.2953,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001948349055887029,
+      "loss": 6.1932,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019478909137505112,
+      "loss": 5.7758,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000194743080307037,
+      "loss": 5.7489,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001946968724802146,
+      "loss": 5.1968,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019465046799054656,
+      "loss": 7.0363,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019460386693440394,
+      "loss": 6.8473,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000194557069408566,
+      "loss": 6.4497,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019451007551022013,
+      "loss": 6.1776,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019446288533696144,
+      "loss": 5.9771,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019441549898679261,
+      "loss": 6.0515,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001943679165581239,
+      "loss": 5.414,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001943201381497726,
+      "loss": 5.6219,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019427216386096313,
+      "loss": 5.3474,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019422399379132662,
+      "loss": 6.7194,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019417562804090082,
+      "loss": 5.8294,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019412706671012998,
+      "loss": 5.7155,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001940783098998643,
+      "loss": 4.9434,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001940293577113601,
+      "loss": 5.8938,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001939802102462794,
+      "loss": 6.2217,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019393086760668981,
+      "loss": 6.9913,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019388132989506422,
+      "loss": 5.8959,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019383159721428065,
+      "loss": 4.9565,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019378166966762205,
+      "loss": 7.3789,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019373154735877603,
+      "loss": 6.6415,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019368123039183467,
+      "loss": 6.7264,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019363071887129433,
+      "loss": 5.8799,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019358001290205543,
+      "loss": 5.8346,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019352911258942217,
+      "loss": 7.3592,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019347801803910235,
+      "loss": 6.7213,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001934267293572072,
+      "loss": 6.1556,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019337524665025116,
+      "loss": 7.0338,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019332357002515145,
+      "loss": 7.0254,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019327169958922814,
+      "loss": 6.6414,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019321963545020378,
+      "loss": 6.2759,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019316737771620318,
+      "loss": 5.4899,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001931149264957532,
+      "loss": 7.5467,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019306228189778254,
+      "loss": 6.259,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019300944403162148,
+      "loss": 5.4746,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001929564130070017,
+      "loss": 5.8942,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019290318893405597,
+      "loss": 6.0119,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019284977192331805,
+      "loss": 6.6972,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019279616208572233,
+      "loss": 5.7639,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019274235953260367,
+      "loss": 6.5201,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019268836437569712,
+      "loss": 6.5406,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019263417672713785,
+      "loss": 6.0901,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001925797966994606,
+      "loss": 6.2198,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019252522440559977,
+      "loss": 5.8672,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019247045995888904,
+      "loss": 5.8941,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019241550347306107,
+      "loss": 6.3477,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019236035506224746,
+      "loss": 5.547,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019230501484097826,
+      "loss": 6.629,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019224948292418192,
+      "loss": 5.8993,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019219375942718507,
+      "loss": 6.2941,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019213784446571209,
+      "loss": 6.4917,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019208173815588503,
+      "loss": 5.3008,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019202544061422335,
+      "loss": 4.9127,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019196895195764362,
+      "loss": 5.7709,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019191227230345935,
+      "loss": 6.4728,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019185540176938065,
+      "loss": 6.3637,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019179834047351407,
+      "loss": 5.6557,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019174108853436232,
+      "loss": 5.9318,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001916836460708241,
+      "loss": 6.7529,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019162601320219365,
+      "loss": 5.7877,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019156819004816073,
+      "loss": 6.2231,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001915101767288103,
+      "loss": 7.7989,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019145197336462215,
+      "loss": 7.5015,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019139358007647084,
+      "loss": 6.4936,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001913349969856253,
+      "loss": 5.4726,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019127622421374866,
+      "loss": 6.5733,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019121726188289792,
+      "loss": 5.7433,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019115811011552388,
+      "loss": 5.8587,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019109876903447063,
+      "loss": 5.0999,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019103923876297545,
+      "loss": 5.1808,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019097951942466854,
+      "loss": 5.4587,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019091961114357274,
+      "loss": 6.1684,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019085951404410327,
+      "loss": 5.6806,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001907992282510675,
+      "loss": 6.5915,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019073875388966465,
+      "loss": 5.9276,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001906780910854856,
+      "loss": 6.7559,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019061723996451248,
+      "loss": 5.7748,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019055620065311861,
+      "loss": 7.0909,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019049497327806812,
+      "loss": 6.4986,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019043355796651565,
+      "loss": 5.216,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019037195484600616,
+      "loss": 8.1231,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001903101640444747,
+      "loss": 5.8963,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019024818569024602,
+      "loss": 7.1114,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001901860199120344,
+      "loss": 6.2322,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019012366683894333,
+      "loss": 6.3287,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019006112660046532,
+      "loss": 3.8251,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018999839932648156,
+      "loss": 6.1936,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001899354851472616,
+      "loss": 5.0188,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018987238419346326,
+      "loss": 6.9289,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018980909659613215,
+      "loss": 5.341,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018974562248670157,
+      "loss": 5.3885,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018968196199699208,
+      "loss": 5.8332,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001896181152592114,
+      "loss": 6.3618,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018955408240595395,
+      "loss": 5.903,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018948986357020074,
+      "loss": 4.3046,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018942545888531897,
+      "loss": 5.6697,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018936086848506185,
+      "loss": 6.1557,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018929609250356827,
+      "loss": 6.1232,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018923113107536247,
+      "loss": 5.1882,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018916598433535388,
+      "loss": 6.3709,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001891006524188368,
+      "loss": 5.7597,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018903513546149,
+      "loss": 5.0299,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001889694335993766,
+      "loss": 5.4972,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018890354696894375,
+      "loss": 6.1881,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018883747570702224,
+      "loss": 5.9239,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001887712199508264,
+      "loss": 6.5558,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001887047798379536,
+      "loss": 6.4772,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018863815550638414,
+      "loss": 6.3761,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018857134709448093,
+      "loss": 6.7836,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018850435474098903,
+      "loss": 5.6181,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018843717858503571,
+      "loss": 6.8446,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001883698187661298,
+      "loss": 6.0209,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018830227542416153,
+      "loss": 6.4907,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018823454869940242,
+      "loss": 6.6526,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018816663873250474,
+      "loss": 6.5424,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00018809854566450128,
+      "loss": 5.7829,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001880302696368051,
+      "loss": 6.267,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001879618107912093,
+      "loss": 5.7366,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018789316926988656,
+      "loss": 5.7723,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000187824345215389,
+      "loss": 5.7909,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001877553387706477,
+      "loss": 5.0467,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018768615007897275,
+      "loss": 5.4952,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018761677928405242,
+      "loss": 6.1704,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018754722652995347,
+      "loss": 5.696,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018747749196112033,
+      "loss": 6.8244,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018740757572237508,
+      "loss": 5.9158,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018733747795891709,
+      "loss": 6.5294,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018726719881632277,
+      "loss": 5.9328,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001871967384405451,
+      "loss": 7.0029,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018712609697791352,
+      "loss": 5.2844,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001870552745751335,
+      "loss": 6.2338,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018698427137928635,
+      "loss": 5.9738,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018691308753782874,
+      "loss": 6.3319,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001868417231985926,
+      "loss": 6.3114,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018677017850978462,
+      "loss": 6.3429,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018669845361998606,
+      "loss": 5.3071,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018662654867815248,
+      "loss": 5.4354,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001865544638336133,
+      "loss": 5.6869,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018648219923607156,
+      "loss": 5.2612,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001864097550356036,
+      "loss": 6.2815,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018633713138265884,
+      "loss": 5.1415,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001862643284280592,
+      "loss": 6.0049,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018619134632299913,
+      "loss": 7.1914,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018611818521904508,
+      "loss": 5.944,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018604484526813523,
+      "loss": 6.1174,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001859713266225792,
+      "loss": 5.0273,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018589762943505765,
+      "loss": 7.1072,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018582375385862217,
+      "loss": 6.9451,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018574970004669466,
+      "loss": 6.883,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018567546815306726,
+      "loss": 5.7982,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018560105833190194,
+      "loss": 5.4688,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001855264707377302,
+      "loss": 6.2033,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018545170552545265,
+      "loss": 5.8918,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018537676285033887,
+      "loss": 6.3683,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018530164286802693,
+      "loss": 6.6208,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018522634573452313,
+      "loss": 5.1474,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018515087160620168,
+      "loss": 7.3018,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018507522063980435,
+      "loss": 6.0014,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018499939299244025,
+      "loss": 5.8694,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018492338882158523,
+      "loss": 6.4814,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018484720828508192,
+      "loss": 5.2886,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018477085154113912,
+      "loss": 5.4669,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018469431874833166,
+      "loss": 5.8911,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001846176100655998,
+      "loss": 5.5574,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001845407256522493,
+      "loss": 6.7941,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018446366566795075,
+      "loss": 6.7376,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018438643027273937,
+      "loss": 5.3374,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018430901962701465,
+      "loss": 6.8929,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001842314338915401,
+      "loss": 6.3781,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018415367322744278,
+      "loss": 5.5445,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001840757377962131,
+      "loss": 6.5702,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018399762775970436,
+      "loss": 5.3821,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001839193432801325,
+      "loss": 6.4985,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018384088452007578,
+      "loss": 5.7335,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018376225164247427,
+      "loss": 6.2381,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018368344481062979,
+      "loss": 6.1422,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018360446418820532,
+      "loss": 6.1602,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018352530993922484,
+      "loss": 6.3867,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018344598222807282,
+      "loss": 5.6421,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018336648121949404,
+      "loss": 6.7861,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018328680707859317,
+      "loss": 6.5187,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018320695997083443,
+      "loss": 6.3333,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001831269400620412,
+      "loss": 7.706,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001830467475183958,
+      "loss": 5.1925,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018296638250643902,
+      "loss": 5.437,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018288584519306986,
+      "loss": 6.165,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018280513574554514,
+      "loss": 5.2935,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018272425433147916,
+      "loss": 6.1547,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001826432011188433,
+      "loss": 6.07,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018256197627596583,
+      "loss": 6.2525,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018248057997153136,
+      "loss": 5.6492,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018239901237458064,
+      "loss": 4.5504,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018231727365451017,
+      "loss": 6.0539,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018223536398107176,
+      "loss": 6.6637,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001821532835243723,
+      "loss": 5.0308,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018207103245487336,
+      "loss": 5.4866,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018198861094339077,
+      "loss": 5.7032,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018190601916109444,
+      "loss": 6.811,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018182325727950777,
+      "loss": 6.0294,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018174032547050754,
+      "loss": 5.3047,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001816572239063233,
+      "loss": 6.3424,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001815739527595372,
+      "loss": 6.0888,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018149051220308363,
+      "loss": 6.9427,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018140690241024872,
+      "loss": 5.676,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018132312355467005,
+      "loss": 5.7181,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001812391758103364,
+      "loss": 5.7546,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018115505935158724,
+      "loss": 5.5083,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001810707743531124,
+      "loss": 6.5114,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018098632098995175,
+      "loss": 6.0489,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018090169943749476,
+      "loss": 5.9063,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001808169098714803,
+      "loss": 7.6832,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018073195246799605,
+      "loss": 6.5052,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018064682740347835,
+      "loss": 5.1272,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018056153485471166,
+      "loss": 6.5013,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018047607499882827,
+      "loss": 6.7378,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018039044801330793,
+      "loss": 5.7988,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018030465407597754,
+      "loss": 5.0814,
+      "step": 558
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018021869336501064,
+      "loss": 6.3901,
+      "step": 559
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018013256605892716,
+      "loss": 4.6266,
+      "step": 560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000180046272336593,
+      "loss": 6.2804,
+      "step": 561
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017995981237721963,
+      "loss": 5.4927,
+      "step": 562
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017987318636036382,
+      "loss": 6.3714,
+      "step": 563
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017978639446592718,
+      "loss": 6.0735,
+      "step": 564
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017969943687415576,
+      "loss": 4.6868,
+      "step": 565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001796123137656398,
+      "loss": 5.6827,
+      "step": 566
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017952502532131326,
+      "loss": 6.1295,
+      "step": 567
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017943757172245338,
+      "loss": 4.8618,
+      "step": 568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017934995315068051,
+      "loss": 6.1436,
+      "step": 569
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001792621697879575,
+      "loss": 5.7896,
+      "step": 570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017917422181658953,
+      "loss": 4.8362,
+      "step": 571
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017908610941922359,
+      "loss": 4.7758,
+      "step": 572
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001789978327788481,
+      "loss": 6.4653,
+      "step": 573
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017890939207879263,
+      "loss": 5.3714,
+      "step": 574
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017882078750272742,
+      "loss": 6.7952,
+      "step": 575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017873201923466315,
+      "loss": 5.9796,
+      "step": 576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017864308745895024,
+      "loss": 6.1053,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001785539923602789,
+      "loss": 7.285,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017846473412367844,
+      "loss": 6.2972,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017837531293451688,
+      "loss": 6.1528,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017828572897850075,
+      "loss": 5.933,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017819598244167461,
+      "loss": 5.6527,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001781060735104206,
+      "loss": 4.9627,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017801600237145823,
+      "loss": 4.909,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017792576921184372,
+      "loss": 6.0909,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017783537421896988,
+      "loss": 6.7453,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017774481758056553,
+      "loss": 5.2949,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017765409948469532,
+      "loss": 5.0662,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017756322011975906,
+      "loss": 6.2726,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001774721796744916,
+      "loss": 5.9916,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017738097833796217,
+      "loss": 6.2743,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001772896162995743,
+      "loss": 5.9074,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017719809374906512,
+      "loss": 7.0059,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017710641087650522,
+      "loss": 5.7142,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017701456787229804,
+      "loss": 6.2126,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001769225649271797,
+      "loss": 6.9569,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017683040223221826,
+      "loss": 5.5663,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017673807997881382,
+      "loss": 5.7133,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017664559835869762,
+      "loss": 6.6871,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017655295756393202,
+      "loss": 5.7252,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017646015778690984,
+      "loss": 6.2929,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001763671992203541,
+      "loss": 5.4647,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017627408205731762,
+      "loss": 6.4771,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017618080649118255,
+      "loss": 5.4968,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017608737271566005,
+      "loss": 6.1935,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001759937809247898,
+      "loss": 5.5003,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017590003131293967,
+      "loss": 5.4991,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017580612407480523,
+      "loss": 6.2404,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017571205940540945,
+      "loss": 5.4565,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001756178375001023,
+      "loss": 5.7853,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017552345855456015,
+      "loss": 7.3182,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017542892276478563,
+      "loss": 6.076,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017533423032710705,
+      "loss": 6.9296,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017523938143817807,
+      "loss": 6.0482,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001751443762949772,
+      "loss": 6.0851,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017504921509480748,
+      "loss": 5.5827,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017495389803529613,
+      "loss": 6.5108,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001748584253143939,
+      "loss": 5.7035,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001747627971303749,
+      "loss": 6.1776,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017466701368183611,
+      "loss": 5.8519,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017457107516769696,
+      "loss": 7.1918,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001744749817871988,
+      "loss": 4.0434,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017437873373990476,
+      "loss": 6.2096,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001742823312256991,
+      "loss": 5.6962,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017418577444478688,
+      "loss": 6.997,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001740890635976935,
+      "loss": 5.8597,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017399219888526436,
+      "loss": 7.7614,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017389518050866444,
+      "loss": 5.3799,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017379800866937773,
+      "loss": 6.1705,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001737006835692071,
+      "loss": 4.9479,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001736032054102734,
+      "loss": 5.4134,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017350557439501575,
+      "loss": 5.6661,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017340779072619043,
+      "loss": 5.5888,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001733098546068708,
+      "loss": 5.9749,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017321176624044687,
+      "loss": 5.5373,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017311352583062483,
+      "loss": 7.0158,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001730151335814266,
+      "loss": 4.8749,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017291658969718942,
+      "loss": 5.7816,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017281789438256548,
+      "loss": 6.4951,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017271904784252145,
+      "loss": 6.2199,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017262005028233804,
+      "loss": 5.0393,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017252090190760953,
+      "loss": 5.7559,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017242160292424362,
+      "loss": 4.3468,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017232215353846052,
+      "loss": 5.8335,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017222255395679296,
+      "loss": 6.612,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017212280438608555,
+      "loss": 5.2109,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017202290503349436,
+      "loss": 6.028,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001719228561064866,
+      "loss": 5.0922,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017182265781284,
+      "loss": 6.1625,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017172231036064253,
+      "loss": 6.03,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017162181395829205,
+      "loss": 5.8901,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017152116881449553,
+      "loss": 6.3785,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017142037513826904,
+      "loss": 5.7265,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017131943313893698,
+      "loss": 6.1871,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017121834302613186,
+      "loss": 6.0202,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017111710500979378,
+      "loss": 5.9053,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017101571930016997,
+      "loss": 4.8746,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001709141861078144,
+      "loss": 4.8643,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017081250564358728,
+      "loss": 4.8514,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017071067811865476,
+      "loss": 5.4811,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017060870374448835,
+      "loss": 4.8295,
+      "step": 661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017050658273286452,
+      "loss": 7.4367,
+      "step": 662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017040431529586426,
+      "loss": 5.5414,
+      "step": 663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017030190164587277,
+      "loss": 5.8422,
+      "step": 664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017019934199557867,
+      "loss": 6.4211,
+      "step": 665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000170096636557974,
+      "loss": 5.5543,
+      "step": 666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001699937855463535,
+      "loss": 5.4912,
+      "step": 667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016989078917431418,
+      "loss": 6.1163,
+      "step": 668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000169787647655755,
+      "loss": 6.2628,
+      "step": 669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001696843612048763,
+      "loss": 6.5464,
+      "step": 670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016958093003617943,
+      "loss": 5.6012,
+      "step": 671
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016947735436446634,
+      "loss": 5.4413,
+      "step": 672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016937363440483894,
+      "loss": 5.5129,
+      "step": 673
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016926977037269895,
+      "loss": 6.1921,
+      "step": 674
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016916576248374719,
+      "loss": 6.5515,
+      "step": 675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016906161095398329,
+      "loss": 6.1102,
+      "step": 676
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016895731599970515,
+      "loss": 6.063,
+      "step": 677
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016885287783750856,
+      "loss": 6.1072,
+      "step": 678
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016874829668428665,
+      "loss": 5.3247,
+      "step": 679
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016864357275722967,
+      "loss": 6.2058,
+      "step": 680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016853870627382415,
+      "loss": 5.4334,
+      "step": 681
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016843369745185287,
+      "loss": 6.2944,
+      "step": 682
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001683285465093941,
+      "loss": 6.8036,
+      "step": 683
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016822325366482136,
+      "loss": 6.4155,
+      "step": 684
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001681178191368027,
+      "loss": 4.6616,
+      "step": 685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001680122431443006,
+      "loss": 5.2746,
+      "step": 686
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016790652590657125,
+      "loss": 6.1679,
+      "step": 687
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016780066764316413,
+      "loss": 6.5925,
+      "step": 688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016769466857392167,
+      "loss": 5.7936,
+      "step": 689
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016758852891897864,
+      "loss": 6.3905,
+      "step": 690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016748224889876186,
+      "loss": 4.7813,
+      "step": 691
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016737582873398963,
+      "loss": 5.456,
+      "step": 692
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001672692686456712,
+      "loss": 4.9898,
+      "step": 693
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016716256885510657,
+      "loss": 6.1624,
+      "step": 694
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016705572958388576,
+      "loss": 5.8066,
+      "step": 695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016694875105388845,
+      "loss": 5.3742,
+      "step": 696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016684163348728364,
+      "loss": 5.5225,
+      "step": 697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016673437710652893,
+      "loss": 5.8621,
+      "step": 698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016662698213437026,
+      "loss": 5.2078,
+      "step": 699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016651944879384148,
+      "loss": 5.3788,
+      "step": 700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016641177730826367,
+      "loss": 5.0009,
+      "step": 701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001663039679012449,
+      "loss": 5.1532,
+      "step": 702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016619602079667956,
+      "loss": 6.5906,
+      "step": 703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016608793621874813,
+      "loss": 6.8414,
+      "step": 704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001659797143919165,
+      "loss": 6.0827,
+      "step": 705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001658713555409356,
+      "loss": 5.5933,
+      "step": 706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000165762859890841,
+      "loss": 5.7127,
+      "step": 707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016565422766695232,
+      "loss": 5.8037,
+      "step": 708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016554545909487276,
+      "loss": 5.1671,
+      "step": 709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016543655440048877,
+      "loss": 5.7564,
+      "step": 710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016532751380996948,
+      "loss": 6.4035,
+      "step": 711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016521833754976615,
+      "loss": 5.3353,
+      "step": 712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016510902584661194,
+      "loss": 4.2235,
+      "step": 713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001649995789275212,
+      "loss": 5.9706,
+      "step": 714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016488999701978903,
+      "loss": 5.6697,
+      "step": 715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016478028035099107,
+      "loss": 6.5164,
+      "step": 716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016467042914898263,
+      "loss": 6.0842,
+      "step": 717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001645604436418985,
+      "loss": 5.5249,
+      "step": 718
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001644503240581524,
+      "loss": 5.4099,
+      "step": 719
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001643400706264364,
+      "loss": 5.027,
+      "step": 720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016422968357572074,
+      "loss": 5.5105,
+      "step": 721
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016411916313525292,
+      "loss": 5.745,
+      "step": 722
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001640085095345576,
+      "loss": 5.9648,
+      "step": 723
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000163897723003436,
+      "loss": 5.9415,
+      "step": 724
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016378680377196524,
+      "loss": 4.748,
+      "step": 725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016367575207049825,
+      "loss": 4.8942,
+      "step": 726
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016356456812966288,
+      "loss": 5.0619,
+      "step": 727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016345325218036173,
+      "loss": 6.1303,
+      "step": 728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016334180445377156,
+      "loss": 5.8531,
+      "step": 729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016323022518134264,
+      "loss": 6.8459,
+      "step": 730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016311851459479862,
+      "loss": 6.7258,
+      "step": 731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016300667292613573,
+      "loss": 5.3146,
+      "step": 732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001628947004076225,
+      "loss": 5.6304,
+      "step": 733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001627825972717992,
+      "loss": 6.6852,
+      "step": 734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016267036375147725,
+      "loss": 5.4385,
+      "step": 735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016255800007973902,
+      "loss": 5.6156,
+      "step": 736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001624455064899371,
+      "loss": 6.3258,
+      "step": 737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001623328832156938,
+      "loss": 5.9256,
+      "step": 738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016222013049090086,
+      "loss": 4.926,
+      "step": 739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016210724854971885,
+      "loss": 6.6747,
+      "step": 740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001619942376265767,
+      "loss": 5.5316,
+      "step": 741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016188109795617112,
+      "loss": 5.5596,
+      "step": 742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016176782977346624,
+      "loss": 6.3052,
+      "step": 743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016165443331369318,
+      "loss": 4.9965,
+      "step": 744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001615409088123493,
+      "loss": 6.2782,
+      "step": 745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016142725650519795,
+      "loss": 5.7394,
+      "step": 746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016131347662826792,
+      "loss": 6.0681,
+      "step": 747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001611995694178529,
+      "loss": 6.0154,
+      "step": 748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016108553511051104,
+      "loss": 6.0204,
+      "step": 749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016097137394306446,
+      "loss": 6.1986,
+      "step": 750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016085708615259863,
+      "loss": 6.4481,
+      "step": 751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016074267197646212,
+      "loss": 6.0388,
+      "step": 752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016062813165226595,
+      "loss": 4.7861,
+      "step": 753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016051346541788298,
+      "loss": 5.8037,
+      "step": 754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016039867351144778,
+      "loss": 5.0505,
+      "step": 755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016028375617135573,
+      "loss": 6.1704,
+      "step": 756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016016871363626277,
+      "loss": 4.8732,
+      "step": 757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016005354614508487,
+      "loss": 5.8258,
+      "step": 758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015993825393699745,
+      "loss": 4.7755,
+      "step": 759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000159822837251435,
+      "loss": 4.6756,
+      "step": 760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015970729632809044,
+      "loss": 5.6683,
+      "step": 761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015959163140691473,
+      "loss": 5.2032,
+      "step": 762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001594758427281164,
+      "loss": 5.8912,
+      "step": 763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001593599305321609,
+      "loss": 4.525,
+      "step": 764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015924389505977038,
+      "loss": 6.1367,
+      "step": 765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015912773655192273,
+      "loss": 5.2768,
+      "step": 766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015901145524985151,
+      "loss": 5.094,
+      "step": 767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015889505139504535,
+      "loss": 5.6614,
+      "step": 768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015877852522924732,
+      "loss": 5.9944,
+      "step": 769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015866187699445448,
+      "loss": 5.5876,
+      "step": 770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015854510693291744,
+      "loss": 4.9416,
+      "step": 771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015842821528713977,
+      "loss": 5.6723,
+      "step": 772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015831120229987762,
+      "loss": 5.0313,
+      "step": 773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015819406821413907,
+      "loss": 5.842,
+      "step": 774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001580768132731837,
+      "loss": 5.7957,
+      "step": 775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001579594377205221,
+      "loss": 6.4825,
+      "step": 776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015784194179991533,
+      "loss": 5.4543,
+      "step": 777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015772432575537445,
+      "loss": 5.9534,
+      "step": 778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015760658983115993,
+      "loss": 4.6472,
+      "step": 779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015748873427178127,
+      "loss": 5.6024,
+      "step": 780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015737075932199636,
+      "loss": 5.4321,
+      "step": 781
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015725266522681111,
+      "loss": 4.9037,
+      "step": 782
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015713445223147876,
+      "loss": 4.8209,
+      "step": 783
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001570161205814996,
+      "loss": 5.541,
+      "step": 784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015689767052262026,
+      "loss": 5.5062,
+      "step": 785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015677910230083326,
+      "loss": 5.8113,
+      "step": 786
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001566604161623766,
+      "loss": 5.0383,
+      "step": 787
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001565416123537331,
+      "loss": 6.019,
+      "step": 788
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015642269112162998,
+      "loss": 6.6402,
+      "step": 789
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015630365271303827,
+      "loss": 4.4452,
+      "step": 790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015618449737517243,
+      "loss": 5.8528,
+      "step": 791
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015606522535548969,
+      "loss": 5.7144,
+      "step": 792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015594583690168966,
+      "loss": 5.1069,
+      "step": 793
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015582633226171366,
+      "loss": 5.2682,
+      "step": 794
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015570671168374438,
+      "loss": 5.8389,
+      "step": 795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015558697541620528,
+      "loss": 5.8461,
+      "step": 796
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015546712370776005,
+      "loss": 8.1143,
+      "step": 797
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015534715680731219,
+      "loss": 6.039,
+      "step": 798
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015522707496400427,
+      "loss": 4.7389,
+      "step": 799
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015510687842721773,
+      "loss": 6.5297,
+      "step": 800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001549865674465722,
+      "loss": 4.9452,
+      "step": 801
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015486614227192482,
+      "loss": 5.0314,
+      "step": 802
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015474560315337005,
+      "loss": 5.4622,
+      "step": 803
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001546249503412389,
+      "loss": 6.6215,
+      "step": 804
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001545041840860986,
+      "loss": 6.079,
+      "step": 805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001543833046387518,
+      "loss": 5.2794,
+      "step": 806
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001542623122502363,
+      "loss": 6.6348,
+      "step": 807
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015414120717182454,
+      "loss": 4.7772,
+      "step": 808
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015401998965502292,
+      "loss": 6.602,
+      "step": 809
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015389865995157128,
+      "loss": 5.6499,
+      "step": 810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015377721831344256,
+      "loss": 6.1383,
+      "step": 811
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001536556649928421,
+      "loss": 6.9105,
+      "step": 812
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001535340002422072,
+      "loss": 6.7051,
+      "step": 813
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015341222431420657,
+      "loss": 5.2171,
+      "step": 814
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015329033746173975,
+      "loss": 5.6989,
+      "step": 815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015316833993793676,
+      "loss": 4.4367,
+      "step": 816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015304623199615736,
+      "loss": 5.6514,
+      "step": 817
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001529240138899907,
+      "loss": 5.9793,
+      "step": 818
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001528016858732546,
+      "loss": 5.4977,
+      "step": 819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015267924819999527,
+      "loss": 4.7439,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015255670112448657,
+      "loss": 4.7903,
+      "step": 821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015243404490122958,
+      "loss": 5.4015,
+      "step": 822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015231127978495207,
+      "loss": 4.948,
+      "step": 823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015218840603060794,
+      "loss": 4.9378,
+      "step": 824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001520654238933767,
+      "loss": 4.9082,
+      "step": 825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015194233362866293,
+      "loss": 5.9205,
+      "step": 826
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015181913549209582,
+      "loss": 6.0312,
+      "step": 827
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015169582973952853,
+      "loss": 4.4628,
+      "step": 828
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015157241662703777,
+      "loss": 6.0409,
+      "step": 829
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015144889641092313,
+      "loss": 6.1643,
+      "step": 830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001513252693477067,
+      "loss": 5.4137,
+      "step": 831
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015120153569413242,
+      "loss": 5.7999,
+      "step": 832
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015107769570716565,
+      "loss": 7.1988,
+      "step": 833
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015095374964399253,
+      "loss": 4.8674,
+      "step": 834
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015082969776201947,
+      "loss": 5.2353,
+      "step": 835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001507055403188727,
+      "loss": 6.5194,
+      "step": 836
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001505812775723977,
+      "loss": 5.9937,
+      "step": 837
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001504569097806585,
+      "loss": 5.0011,
+      "step": 838
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015033243720193745,
+      "loss": 5.4593,
+      "step": 839
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015020786009473444,
+      "loss": 5.0385,
+      "step": 840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00015008317871776644,
+      "loss": 5.938,
+      "step": 841
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001499583933299669,
+      "loss": 6.054,
+      "step": 842
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014983350419048544,
+      "loss": 6.7018,
+      "step": 843
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014970851155868698,
+      "loss": 5.1695,
+      "step": 844
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014958341569415147,
+      "loss": 6.2491,
+      "step": 845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014945821685667326,
+      "loss": 5.9977,
+      "step": 846
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014933291530626046,
+      "loss": 6.6758,
+      "step": 847
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014920751130313454,
+      "loss": 5.0341,
+      "step": 848
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014908200510772978,
+      "loss": 5.9516,
+      "step": 849
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001489563969806926,
+      "loss": 5.5977,
+      "step": 850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001488306871828812,
+      "loss": 6.2267,
+      "step": 851
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014870487597536485,
+      "loss": 5.5812,
+      "step": 852
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001485789636194235,
+      "loss": 5.6439,
+      "step": 853
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014845295037654708,
+      "loss": 5.6932,
+      "step": 854
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001483268365084351,
+      "loss": 7.2018,
+      "step": 855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000148200622276996,
+      "loss": 6.1214,
+      "step": 856
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001480743079443467,
+      "loss": 5.3886,
+      "step": 857
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014794789377281192,
+      "loss": 5.3382,
+      "step": 858
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014782138002492385,
+      "loss": 7.078,
+      "step": 859
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014769476696342137,
+      "loss": 5.6366,
+      "step": 860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001475680548512497,
+      "loss": 6.3989,
+      "step": 861
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014744124395155964,
+      "loss": 6.1227,
+      "step": 862
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014731433452770722,
+      "loss": 5.2983,
+      "step": 863
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014718732684325313,
+      "loss": 5.6128,
+      "step": 864
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014706022116196208,
+      "loss": 5.9942,
+      "step": 865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014693301774780225,
+      "loss": 6.0176,
+      "step": 866
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001468057168649448,
+      "loss": 5.5153,
+      "step": 867
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014667831877776343,
+      "loss": 5.4345,
+      "step": 868
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001465508237508336,
+      "loss": 6.6986,
+      "step": 869
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014642323204893203,
+      "loss": 5.9937,
+      "step": 870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014629554393703634,
+      "loss": 5.8378,
+      "step": 871
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014616775968032434,
+      "loss": 5.442,
+      "step": 872
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014603987954417348,
+      "loss": 6.0183,
+      "step": 873
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014591190379416034,
+      "loss": 6.1252,
+      "step": 874
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014578383269606005,
+      "loss": 4.7206,
+      "step": 875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014565566651584576,
+      "loss": 6.0504,
+      "step": 876
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014552740551968815,
+      "loss": 4.5254,
+      "step": 877
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014539904997395468,
+      "loss": 4.5177,
+      "step": 878
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014527060014520932,
+      "loss": 5.9356,
+      "step": 879
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014514205630021169,
+      "loss": 4.9303,
+      "step": 880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014501341870591682,
+      "loss": 4.9991,
+      "step": 881
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014488468762947427,
+      "loss": 6.2362,
+      "step": 882
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014475586333822786,
+      "loss": 5.0713,
+      "step": 883
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014462694609971495,
+      "loss": 5.4034,
+      "step": 884
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014449793618166593,
+      "loss": 5.7244,
+      "step": 885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014436883385200366,
+      "loss": 6.2117,
+      "step": 886
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014423963937884292,
+      "loss": 4.6413,
+      "step": 887
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014411035303048988,
+      "loss": 5.8028,
+      "step": 888
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001439809750754415,
+      "loss": 5.0245,
+      "step": 889
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001438515057823849,
+      "loss": 6.4042,
+      "step": 890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014372194542019705,
+      "loss": 5.7846,
+      "step": 891
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001435922942579439,
+      "loss": 5.4107,
+      "step": 892
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014346255256488002,
+      "loss": 5.298,
+      "step": 893
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014333272061044802,
+      "loss": 6.4555,
+      "step": 894
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014320279866427796,
+      "loss": 5.378,
+      "step": 895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014307278699618678,
+      "loss": 5.3475,
+      "step": 896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001429426858761777,
+      "loss": 5.047,
+      "step": 897
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014281249557443978,
+      "loss": 6.8231,
+      "step": 898
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014268221636134723,
+      "loss": 4.9412,
+      "step": 899
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000142551848507459,
+      "loss": 6.0637,
+      "step": 900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000142421392283518,
+      "loss": 6.1364,
+      "step": 901
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001422908479604508,
+      "loss": 5.964,
+      "step": 902
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001421602158093668,
+      "loss": 5.1714,
+      "step": 903
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001420294961015579,
+      "loss": 5.4955,
+      "step": 904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014189868910849778,
+      "loss": 6.8632,
+      "step": 905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014176779510184138,
+      "loss": 5.0757,
+      "step": 906
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001416368143534244,
+      "loss": 5.5307,
+      "step": 907
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014150574713526267,
+      "loss": 4.1804,
+      "step": 908
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014137459371955151,
+      "loss": 5.7484,
+      "step": 909
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001412433543786654,
+      "loss": 5.8138,
+      "step": 910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001411120293851571,
+      "loss": 4.1454,
+      "step": 911
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014098061901175748,
+      "loss": 5.4737,
+      "step": 912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014084912353137444,
+      "loss": 5.6166,
+      "step": 913
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014071754321709284,
+      "loss": 6.3713,
+      "step": 914
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014058587834217355,
+      "loss": 5.8209,
+      "step": 915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014045412918005333,
+      "loss": 6.2799,
+      "step": 916
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014032229600434364,
+      "loss": 6.5622,
+      "step": 917
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001401903790888307,
+      "loss": 4.7806,
+      "step": 918
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00014005837870747438,
+      "loss": 6.941,
+      "step": 919
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001399262951344082,
+      "loss": 4.8503,
+      "step": 920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001397941286439381,
+      "loss": 5.0633,
+      "step": 921
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013966187951054254,
+      "loss": 4.5215,
+      "step": 922
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013952954800887138,
+      "loss": 4.7077,
+      "step": 923
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013939713441374568,
+      "loss": 5.1951,
+      "step": 924
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001392646390001569,
+      "loss": 4.8272,
+      "step": 925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013913206204326645,
+      "loss": 5.2775,
+      "step": 926
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001389994038184051,
+      "loss": 6.0271,
+      "step": 927
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013886666460107236,
+      "loss": 6.0238,
+      "step": 928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000138733844666936,
+      "loss": 5.79,
+      "step": 929
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013860094429183137,
+      "loss": 4.6075,
+      "step": 930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013846796375176082,
+      "loss": 5.6981,
+      "step": 931
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001383349033228934,
+      "loss": 5.469,
+      "step": 932
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001382017632815638,
+      "loss": 4.9522,
+      "step": 933
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013806854390427224,
+      "loss": 5.9167,
+      "step": 934
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013793524546768356,
+      "loss": 6.8896,
+      "step": 935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013780186824862694,
+      "loss": 5.7714,
+      "step": 936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000137668412524095,
+      "loss": 5.8282,
+      "step": 937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013753487857124362,
+      "loss": 4.6471,
+      "step": 938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013740126666739085,
+      "loss": 6.4687,
+      "step": 939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001372675770900169,
+      "loss": 5.8107,
+      "step": 940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013713381011676315,
+      "loss": 5.889,
+      "step": 941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001369999660254317,
+      "loss": 5.3122,
+      "step": 942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001368660450939849,
+      "loss": 6.5774,
+      "step": 943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013673204760054458,
+      "loss": 5.8088,
+      "step": 944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001365979738233916,
+      "loss": 6.3982,
+      "step": 945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013646382404096534,
+      "loss": 5.8301,
+      "step": 946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001363295985318628,
+      "loss": 4.4446,
+      "step": 947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013619529757483856,
+      "loss": 4.8547,
+      "step": 948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001360609214488036,
+      "loss": 5.4619,
+      "step": 949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013592647043282513,
+      "loss": 5.4053,
+      "step": 950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001357919448061259,
+      "loss": 4.6703,
+      "step": 951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013565734484808362,
+      "loss": 5.5948,
+      "step": 952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013552267083823027,
+      "loss": 4.6342,
+      "step": 953
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013538792305625176,
+      "loss": 5.931,
+      "step": 954
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013525310178198705,
+      "loss": 4.8818,
+      "step": 955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001351182072954279,
+      "loss": 5.9521,
+      "step": 956
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013498323987671802,
+      "loss": 5.0234,
+      "step": 957
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013484819980615252,
+      "loss": 4.9403,
+      "step": 958
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013471308736417748,
+      "loss": 3.0792,
+      "step": 959
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013457790283138928,
+      "loss": 5.8687,
+      "step": 960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013444264648853396,
+      "loss": 4.6002,
+      "step": 961
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013430731861650673,
+      "loss": 5.1752,
+      "step": 962
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013417191949635136,
+      "loss": 5.4423,
+      "step": 963
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001340364494092596,
+      "loss": 3.9228,
+      "step": 964
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013390090863657047,
+      "loss": 5.7721,
+      "step": 965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013376529745976994,
+      "loss": 5.5789,
+      "step": 966
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013362961616049005,
+      "loss": 5.4092,
+      "step": 967
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013349386502050858,
+      "loss": 5.7409,
+      "step": 968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001333580443217483,
+      "loss": 4.6577,
+      "step": 969
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013322215434627646,
+      "loss": 3.875,
+      "step": 970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013308619537630414,
+      "loss": 5.2999,
+      "step": 971
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013295016769418577,
+      "loss": 4.9812,
+      "step": 972
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001328140715824184,
+      "loss": 6.2058,
+      "step": 973
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013267790732364129,
+      "loss": 5.6329,
+      "step": 974
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013254167520063509,
+      "loss": 4.8895,
+      "step": 975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013240537549632153,
+      "loss": 4.8129,
+      "step": 976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013226900849376262,
+      "loss": 6.2137,
+      "step": 977
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001321325744761601,
+      "loss": 5.0612,
+      "step": 978
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013199607372685497,
+      "loss": 4.9363,
+      "step": 979
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013185950652932671,
+      "loss": 5.9166,
+      "step": 980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013172287316719295,
+      "loss": 4.5175,
+      "step": 981
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013158617392420856,
+      "loss": 5.5664,
+      "step": 982
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001314494090842653,
+      "loss": 4.968,
+      "step": 983
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013131257893139118,
+      "loss": 6.2302,
+      "step": 984
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013117568374974988,
+      "loss": 5.6751,
+      "step": 985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013103872382364,
+      "loss": 5.5975,
+      "step": 986
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013090169943749476,
+      "loss": 5.1843,
+      "step": 987
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001307646108758811,
+      "loss": 6.778,
+      "step": 988
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013062745842349935,
+      "loss": 6.0237,
+      "step": 989
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013049024236518243,
+      "loss": 5.2664,
+      "step": 990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013035296298589548,
+      "loss": 5.6303,
+      "step": 991
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000130215620570735,
+      "loss": 5.4345,
+      "step": 992
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00013007821540492854,
+      "loss": 5.0638,
+      "step": 993
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001299407477738338,
+      "loss": 6.2008,
+      "step": 994
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00012980321796293836,
+      "loss": 5.4946,
+      "step": 995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001296656262578589,
+      "loss": 5.8284,
+      "step": 996
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001295279729443406,
+      "loss": 6.1732,
+      "step": 997
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00012939025830825657,
+      "loss": 6.2755,
+      "step": 998
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00012925248263560732,
+      "loss": 5.0263,
+      "step": 999
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001291146462125201,
+      "loss": 4.3647,
+      "step": 1000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00012897674932524833,
+      "loss": 4.7797,
+      "step": 1001
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000128838792260171,
+      "loss": 5.1862,
+      "step": 1002
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00012870077530379204,
+      "loss": 4.3574,
+      "step": 1003
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00012856269874273984,
+      "loss": 5.4475,
+      "step": 1004
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001284245628637665,
+      "loss": 5.7,
+      "step": 1005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001282863679537473,
+      "loss": 5.8256,
+      "step": 1006
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001281481142996802,
+      "loss": 5.8949,
+      "step": 1007
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00012800980218868508,
+      "loss": 5.8213,
+      "step": 1008
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001278714319080033,
+      "loss": 5.3477,
+      "step": 1009
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012773300374499687,
+      "loss": 5.1089,
+      "step": 1010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012759451798714815,
+      "loss": 5.2621,
+      "step": 1011
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012745597492205913,
+      "loss": 6.1217,
+      "step": 1012
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012731737483745068,
+      "loss": 6.2075,
+      "step": 1013
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012717871802116217,
+      "loss": 5.8824,
+      "step": 1014
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001270400047611508,
+      "loss": 4.877,
+      "step": 1015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012690123534549094,
+      "loss": 5.9529,
+      "step": 1016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012676241006237367,
+      "loss": 5.5714,
+      "step": 1017
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012662352920010593,
+      "loss": 6.1342,
+      "step": 1018
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012648459304711026,
+      "loss": 4.9639,
+      "step": 1019
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012634560189192395,
+      "loss": 5.0054,
+      "step": 1020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001262065560231985,
+      "loss": 5.24,
+      "step": 1021
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012606745572969907,
+      "loss": 5.7127,
+      "step": 1022
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001259283013003038,
+      "loss": 7.8473,
+      "step": 1023
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012578909302400333,
+      "loss": 5.8946,
+      "step": 1024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012564983118990011,
+      "loss": 5.0391,
+      "step": 1025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001255105160872077,
+      "loss": 5.0307,
+      "step": 1026
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012537114800525046,
+      "loss": 4.6024,
+      "step": 1027
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001252317272334627,
+      "loss": 6.2957,
+      "step": 1028
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001250922540613881,
+      "loss": 5.5082,
+      "step": 1029
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012495272877867926,
+      "loss": 5.1336,
+      "step": 1030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012481315167509692,
+      "loss": 4.2794,
+      "step": 1031
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012467352304050948,
+      "loss": 4.8574,
+      "step": 1032
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001245338431648924,
+      "loss": 5.6893,
+      "step": 1033
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012439411233832742,
+      "loss": 6.4444,
+      "step": 1034
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012425433085100224,
+      "loss": 4.853,
+      "step": 1035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001241144989932097,
+      "loss": 4.9955,
+      "step": 1036
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012397461705534728,
+      "loss": 6.0516,
+      "step": 1037
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001238346853279164,
+      "loss": 5.2923,
+      "step": 1038
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012369470410152194,
+      "loss": 5.4886,
+      "step": 1039
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012355467366687156,
+      "loss": 5.4444,
+      "step": 1040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012341459431477515,
+      "loss": 6.2435,
+      "step": 1041
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001232744663361441,
+      "loss": 4.7896,
+      "step": 1042
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012313429002199088,
+      "loss": 5.0261,
+      "step": 1043
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001229940656634283,
+      "loss": 5.7523,
+      "step": 1044
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012285379355166893,
+      "loss": 5.2303,
+      "step": 1045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012271347397802452,
+      "loss": 4.5538,
+      "step": 1046
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001225731072339054,
+      "loss": 5.994,
+      "step": 1047
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012243269361081994,
+      "loss": 5.7812,
+      "step": 1048
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012229223340037363,
+      "loss": 4.8197,
+      "step": 1049
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000122151726894269,
+      "loss": 5.5173,
+      "step": 1050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012201117438430456,
+      "loss": 4.9564,
+      "step": 1051
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012187057616237439,
+      "loss": 5.481,
+      "step": 1052
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012172993252046747,
+      "loss": 6.0993,
+      "step": 1053
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012158924375066716,
+      "loss": 4.1174,
+      "step": 1054
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012144851014515055,
+      "loss": 5.2046,
+      "step": 1055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001213077319961878,
+      "loss": 5.1727,
+      "step": 1056
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012116690959614154,
+      "loss": 5.3923,
+      "step": 1057
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001210260432374664,
+      "loss": 4.6063,
+      "step": 1058
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012088513321270822,
+      "loss": 5.8659,
+      "step": 1059
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012074417981450361,
+      "loss": 5.3295,
+      "step": 1060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012060318333557914,
+      "loss": 4.7035,
+      "step": 1061
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012046214406875093,
+      "loss": 4.787,
+      "step": 1062
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012032106230692393,
+      "loss": 5.5987,
+      "step": 1063
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012017993834309138,
+      "loss": 5.3134,
+      "step": 1064
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00012003877247033411,
+      "loss": 5.9071,
+      "step": 1065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011989756498182002,
+      "loss": 5.1393,
+      "step": 1066
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011975631617080339,
+      "loss": 6.2076,
+      "step": 1067
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011961502633062441,
+      "loss": 4.6879,
+      "step": 1068
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011947369575470838,
+      "loss": 4.5667,
+      "step": 1069
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011933232473656526,
+      "loss": 5.6305,
+      "step": 1070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011919091356978893,
+      "loss": 3.5505,
+      "step": 1071
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001190494625480568,
+      "loss": 5.7761,
+      "step": 1072
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011890797196512884,
+      "loss": 6.4153,
+      "step": 1073
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011876644211484736,
+      "loss": 5.4257,
+      "step": 1074
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011862487329113606,
+      "loss": 5.816,
+      "step": 1075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011848326578799975,
+      "loss": 6.0371,
+      "step": 1076
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011834161989952343,
+      "loss": 6.1908,
+      "step": 1077
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001181999359198719,
+      "loss": 4.6485,
+      "step": 1078
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011805821414328897,
+      "loss": 5.6462,
+      "step": 1079
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001179164548640971,
+      "loss": 4.8954,
+      "step": 1080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001177746583766965,
+      "loss": 5.5028,
+      "step": 1081
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001176328249755647,
+      "loss": 5.4789,
+      "step": 1082
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011749095495525584,
+      "loss": 5.7678,
+      "step": 1083
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011734904861040021,
+      "loss": 5.0771,
+      "step": 1084
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011720710623570351,
+      "loss": 5.5835,
+      "step": 1085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011706512812594616,
+      "loss": 4.3658,
+      "step": 1086
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001169231145759829,
+      "loss": 6.2523,
+      "step": 1087
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001167810658807421,
+      "loss": 5.1988,
+      "step": 1088
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000116638982335225,
+      "loss": 5.981,
+      "step": 1089
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001164968642345053,
+      "loss": 4.8817,
+      "step": 1090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001163547118737284,
+      "loss": 7.1504,
+      "step": 1091
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011621252554811095,
+      "loss": 6.6443,
+      "step": 1092
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011607030555294005,
+      "loss": 5.3829,
+      "step": 1093
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011592805218357276,
+      "loss": 5.059,
+      "step": 1094
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001157857657354354,
+      "loss": 5.4421,
+      "step": 1095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001156434465040231,
+      "loss": 4.861,
+      "step": 1096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011550109478489894,
+      "loss": 5.202,
+      "step": 1097
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011535871087369357,
+      "loss": 4.7842,
+      "step": 1098
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011521629506610438,
+      "loss": 5.0164,
+      "step": 1099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001150738476578952,
+      "loss": 6.1875,
+      "step": 1100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011493136894489527,
+      "loss": 6.257,
+      "step": 1101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011478885922299898,
+      "loss": 6.3744,
+      "step": 1102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011464631878816501,
+      "loss": 4.6964,
+      "step": 1103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011450374793641596,
+      "loss": 4.9191,
+      "step": 1104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011436114696383749,
+      "loss": 4.2655,
+      "step": 1105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011421851616657789,
+      "loss": 5.3668,
+      "step": 1106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011407585584084726,
+      "loss": 4.8177,
+      "step": 1107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011393316628291724,
+      "loss": 4.8671,
+      "step": 1108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011379044778911993,
+      "loss": 5.389,
+      "step": 1109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011364770065584772,
+      "loss": 4.9202,
+      "step": 1110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011350492517955233,
+      "loss": 5.5754,
+      "step": 1111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011336212165674451,
+      "loss": 5.1908,
+      "step": 1112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011321929038399308,
+      "loss": 6.1045,
+      "step": 1113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011307643165792463,
+      "loss": 5.4175,
+      "step": 1114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011293354577522263,
+      "loss": 5.0777,
+      "step": 1115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011279063303262707,
+      "loss": 4.7641,
+      "step": 1116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011264769372693364,
+      "loss": 6.2132,
+      "step": 1117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011250472815499328,
+      "loss": 4.9227,
+      "step": 1118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011236173661371138,
+      "loss": 5.5174,
+      "step": 1119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011221871940004731,
+      "loss": 5.0953,
+      "step": 1120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011207567681101377,
+      "loss": 5.5031,
+      "step": 1121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011193260914367616,
+      "loss": 6.1063,
+      "step": 1122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001117895166951519,
+      "loss": 4.5614,
+      "step": 1123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011164639976260993,
+      "loss": 3.1751,
+      "step": 1124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011150325864327003,
+      "loss": 5.071,
+      "step": 1125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011136009363440222,
+      "loss": 4.413,
+      "step": 1126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011121690503332614,
+      "loss": 4.4425,
+      "step": 1127
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011107369313741031,
+      "loss": 6.5263,
+      "step": 1128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011093045824407186,
+      "loss": 4.1829,
+      "step": 1129
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011078720065077541,
+      "loss": 4.0731,
+      "step": 1130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011064392065503293,
+      "loss": 5.5707,
+      "step": 1131
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011050061855440284,
+      "loss": 3.9901,
+      "step": 1132
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011035729464648945,
+      "loss": 4.3865,
+      "step": 1133
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011021394922894235,
+      "loss": 4.6455,
+      "step": 1134
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00011007058259945584,
+      "loss": 5.177,
+      "step": 1135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010992719505576827,
+      "loss": 5.9807,
+      "step": 1136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010978378689566144,
+      "loss": 3.5182,
+      "step": 1137
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010964035841695988,
+      "loss": 5.4011,
+      "step": 1138
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010949690991753037,
+      "loss": 4.6314,
+      "step": 1139
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010935344169528132,
+      "loss": 5.5408,
+      "step": 1140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010920995404816204,
+      "loss": 5.756,
+      "step": 1141
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010906644727416215,
+      "loss": 4.6561,
+      "step": 1142
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010892292167131106,
+      "loss": 5.0698,
+      "step": 1143
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010877937753767726,
+      "loss": 5.1544,
+      "step": 1144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010863581517136776,
+      "loss": 5.8699,
+      "step": 1145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010849223487052733,
+      "loss": 4.66,
+      "step": 1146
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010834863693333805,
+      "loss": 5.4031,
+      "step": 1147
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010820502165801869,
+      "loss": 4.9401,
+      "step": 1148
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010806138934282393,
+      "loss": 6.4846,
+      "step": 1149
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010791774028604384,
+      "loss": 5.2888,
+      "step": 1150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010777407478600335,
+      "loss": 4.8582,
+      "step": 1151
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010763039314106145,
+      "loss": 4.5254,
+      "step": 1152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010748669564961069,
+      "loss": 5.3522,
+      "step": 1153
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010734298261007653,
+      "loss": 5.0353,
+      "step": 1154
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001071992543209167,
+      "loss": 5.3398,
+      "step": 1155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010705551108062067,
+      "loss": 5.9226,
+      "step": 1156
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010691175318770888,
+      "loss": 6.6512,
+      "step": 1157
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001067679809407322,
+      "loss": 3.6578,
+      "step": 1158
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010662419463827136,
+      "loss": 4.0295,
+      "step": 1159
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010648039457893625,
+      "loss": 5.8887,
+      "step": 1160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010633658106136537,
+      "loss": 5.4471,
+      "step": 1161
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010619275438422511,
+      "loss": 5.8591,
+      "step": 1162
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001060489148462092,
+      "loss": 6.7124,
+      "step": 1163
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010590506274603813,
+      "loss": 5.2377,
+      "step": 1164
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010576119838245844,
+      "loss": 4.5649,
+      "step": 1165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010561732205424211,
+      "loss": 5.427,
+      "step": 1166
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010547343406018601,
+      "loss": 5.9834,
+      "step": 1167
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010532953469911126,
+      "loss": 4.5466,
+      "step": 1168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010518562426986253,
+      "loss": 5.9321,
+      "step": 1169
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001050417030713075,
+      "loss": 6.5737,
+      "step": 1170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010489777140233618,
+      "loss": 5.4241,
+      "step": 1171
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010475382956186044,
+      "loss": 6.3614,
+      "step": 1172
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010460987784881313,
+      "loss": 5.8513,
+      "step": 1173
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010446591656214768,
+      "loss": 4.0672,
+      "step": 1174
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001043219460008374,
+      "loss": 4.4077,
+      "step": 1175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010417796646387483,
+      "loss": 4.7605,
+      "step": 1176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010403397825027122,
+      "loss": 5.312,
+      "step": 1177
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001038899816590557,
+      "loss": 4.592,
+      "step": 1178
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010374597698927495,
+      "loss": 4.6163,
+      "step": 1179
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010360196453999236,
+      "loss": 5.3376,
+      "step": 1180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010345794461028749,
+      "loss": 5.2007,
+      "step": 1181
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010331391749925538,
+      "loss": 5.5301,
+      "step": 1182
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010316988350600607,
+      "loss": 5.85,
+      "step": 1183
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010302584292966384,
+      "loss": 5.6633,
+      "step": 1184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010288179606936666,
+      "loss": 5.6581,
+      "step": 1185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010273774322426553,
+      "loss": 4.9835,
+      "step": 1186
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010259368469352391,
+      "loss": 4.9253,
+      "step": 1187
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010244962077631704,
+      "loss": 4.9379,
+      "step": 1188
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010230555177183137,
+      "loss": 5.787,
+      "step": 1189
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010216147797926388,
+      "loss": 6.793,
+      "step": 1190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010201739969782153,
+      "loss": 5.52,
+      "step": 1191
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001018733172267206,
+      "loss": 5.4136,
+      "step": 1192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00010172923086518606,
+      "loss": 5.7649,
+      "step": 1193
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010158514091245092,
+      "loss": 5.1225,
+      "step": 1194
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010144104766775572,
+      "loss": 3.5225,
+      "step": 1195
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010129695143034778,
+      "loss": 5.4129,
+      "step": 1196
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010115285249948073,
+      "loss": 5.5132,
+      "step": 1197
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010100875117441364,
+      "loss": 4.9903,
+      "step": 1198
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010086464775441065,
+      "loss": 5.9875,
+      "step": 1199
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010072054253874026,
+      "loss": 5.3933,
+      "step": 1200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010057643582667467,
+      "loss": 5.4503,
+      "step": 1201
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010043232791748917,
+      "loss": 4.6217,
+      "step": 1202
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010028821911046159,
+      "loss": 5.7063,
+      "step": 1203
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00010014410970487155,
+      "loss": 5.9816,
+      "step": 1204
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001,
+      "loss": 6.484,
+      "step": 1205
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.985589029512846e-05,
+      "loss": 5.618,
+      "step": 1206
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.971178088953845e-05,
+      "loss": 4.7334,
+      "step": 1207
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956767208251085e-05,
+      "loss": 5.0552,
+      "step": 1208
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.942356417332538e-05,
+      "loss": 5.9861,
+      "step": 1209
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.927945746125975e-05,
+      "loss": 5.082,
+      "step": 1210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.913535224558935e-05,
+      "loss": 6.1912,
+      "step": 1211
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.899124882558638e-05,
+      "loss": 5.7504,
+      "step": 1212
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884714750051928e-05,
+      "loss": 5.3951,
+      "step": 1213
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870304856965223e-05,
+      "loss": 6.0031,
+      "step": 1214
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.85589523322443e-05,
+      "loss": 6.3893,
+      "step": 1215
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.84148590875491e-05,
+      "loss": 4.7015,
+      "step": 1216
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.827076913481399e-05,
+      "loss": 5.222,
+      "step": 1217
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.81266827732794e-05,
+      "loss": 6.5666,
+      "step": 1218
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.798260030217846e-05,
+      "loss": 4.6564,
+      "step": 1219
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.783852202073613e-05,
+      "loss": 4.7684,
+      "step": 1220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.769444822816864e-05,
+      "loss": 4.9094,
+      "step": 1221
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.755037922368299e-05,
+      "loss": 6.3644,
+      "step": 1222
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.740631530647611e-05,
+      "loss": 5.1803,
+      "step": 1223
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.726225677573448e-05,
+      "loss": 5.0683,
+      "step": 1224
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.711820393063338e-05,
+      "loss": 4.3958,
+      "step": 1225
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.697415707033617e-05,
+      "loss": 6.5514,
+      "step": 1226
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.683011649399394e-05,
+      "loss": 4.8808,
+      "step": 1227
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.668608250074463e-05,
+      "loss": 4.4952,
+      "step": 1228
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.654205538971254e-05,
+      "loss": 4.8186,
+      "step": 1229
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.639803546000766e-05,
+      "loss": 4.9381,
+      "step": 1230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.625402301072507e-05,
+      "loss": 5.1544,
+      "step": 1231
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.611001834094434e-05,
+      "loss": 5.3929,
+      "step": 1232
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.596602174972883e-05,
+      "loss": 5.0314,
+      "step": 1233
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.582203353612516e-05,
+      "loss": 5.3557,
+      "step": 1234
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.567805399916259e-05,
+      "loss": 5.199,
+      "step": 1235
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.553408343785233e-05,
+      "loss": 4.1848,
+      "step": 1236
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.539012215118688e-05,
+      "loss": 5.822,
+      "step": 1237
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.52461704381396e-05,
+      "loss": 4.1251,
+      "step": 1238
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.510222859766383e-05,
+      "loss": 5.4311,
+      "step": 1239
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.495829692869255e-05,
+      "loss": 4.955,
+      "step": 1240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.48143757301375e-05,
+      "loss": 4.8922,
+      "step": 1241
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.467046530088875e-05,
+      "loss": 5.1246,
+      "step": 1242
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.452656593981398e-05,
+      "loss": 3.9227,
+      "step": 1243
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.438267794575791e-05,
+      "loss": 5.1576,
+      "step": 1244
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.423880161754158e-05,
+      "loss": 5.8724,
+      "step": 1245
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.40949372539619e-05,
+      "loss": 6.3436,
+      "step": 1246
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.395108515379082e-05,
+      "loss": 4.4699,
+      "step": 1247
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.380724561577494e-05,
+      "loss": 5.1731,
+      "step": 1248
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.366341893863467e-05,
+      "loss": 5.373,
+      "step": 1249
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.351960542106376e-05,
+      "loss": 5.1584,
+      "step": 1250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.337580536172865e-05,
+      "loss": 4.9543,
+      "step": 1251
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.323201905926783e-05,
+      "loss": 4.9889,
+      "step": 1252
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.308824681229115e-05,
+      "loss": 4.807,
+      "step": 1253
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.294448891937936e-05,
+      "loss": 6.0599,
+      "step": 1254
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.28007456790833e-05,
+      "loss": 5.0216,
+      "step": 1255
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.265701738992352e-05,
+      "loss": 5.5035,
+      "step": 1256
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.251330435038935e-05,
+      "loss": 5.8444,
+      "step": 1257
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.236960685893856e-05,
+      "loss": 5.0595,
+      "step": 1258
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.222592521399665e-05,
+      "loss": 5.0501,
+      "step": 1259
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.208225971395618e-05,
+      "loss": 4.8398,
+      "step": 1260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.19386106571761e-05,
+      "loss": 5.3827,
+      "step": 1261
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.179497834198134e-05,
+      "loss": 5.8723,
+      "step": 1262
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.165136306666196e-05,
+      "loss": 5.3389,
+      "step": 1263
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.150776512947272e-05,
+      "loss": 6.0498,
+      "step": 1264
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.136418482863229e-05,
+      "loss": 4.8891,
+      "step": 1265
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.122062246232273e-05,
+      "loss": 6.7454,
+      "step": 1266
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.107707832868896e-05,
+      "loss": 5.6116,
+      "step": 1267
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.093355272583788e-05,
+      "loss": 4.6506,
+      "step": 1268
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.079004595183799e-05,
+      "loss": 5.2075,
+      "step": 1269
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.064655830471871e-05,
+      "loss": 5.912,
+      "step": 1270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.050309008246964e-05,
+      "loss": 4.878,
+      "step": 1271
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.035964158304017e-05,
+      "loss": 4.5355,
+      "step": 1272
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.021621310433861e-05,
+      "loss": 4.0255,
+      "step": 1273
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.007280494423173e-05,
+      "loss": 6.155,
+      "step": 1274
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.992941740054418e-05,
+      "loss": 5.1002,
+      "step": 1275
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.978605077105768e-05,
+      "loss": 4.8257,
+      "step": 1276
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.964270535351058e-05,
+      "loss": 4.6168,
+      "step": 1277
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.94993814455972e-05,
+      "loss": 6.7373,
+      "step": 1278
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.935607934496708e-05,
+      "loss": 4.9193,
+      "step": 1279
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.921279934922461e-05,
+      "loss": 4.7693,
+      "step": 1280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.906954175592819e-05,
+      "loss": 4.9181,
+      "step": 1281
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.892630686258967e-05,
+      "loss": 4.285,
+      "step": 1282
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.87830949666739e-05,
+      "loss": 4.2733,
+      "step": 1283
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.863990636559778e-05,
+      "loss": 4.438,
+      "step": 1284
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.849674135673e-05,
+      "loss": 6.2872,
+      "step": 1285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.83536002373901e-05,
+      "loss": 6.2179,
+      "step": 1286
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.821048330484814e-05,
+      "loss": 5.5103,
+      "step": 1287
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.806739085632388e-05,
+      "loss": 4.2185,
+      "step": 1288
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.792432318898624e-05,
+      "loss": 4.9297,
+      "step": 1289
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.778128059995269e-05,
+      "loss": 5.4866,
+      "step": 1290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.763826338628864e-05,
+      "loss": 5.7636,
+      "step": 1291
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.749527184500673e-05,
+      "loss": 5.05,
+      "step": 1292
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.735230627306637e-05,
+      "loss": 5.5946,
+      "step": 1293
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.720936696737295e-05,
+      "loss": 5.4511,
+      "step": 1294
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.706645422477739e-05,
+      "loss": 4.8852,
+      "step": 1295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.692356834207541e-05,
+      "loss": 6.4843,
+      "step": 1296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.678070961600693e-05,
+      "loss": 4.6133,
+      "step": 1297
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.663787834325547e-05,
+      "loss": 5.8254,
+      "step": 1298
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.649507482044767e-05,
+      "loss": 6.2873,
+      "step": 1299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.635229934415229e-05,
+      "loss": 5.3347,
+      "step": 1300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.620955221088009e-05,
+      "loss": 5.6486,
+      "step": 1301
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.60668337170828e-05,
+      "loss": 5.4503,
+      "step": 1302
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.592414415915275e-05,
+      "loss": 5.7778,
+      "step": 1303
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.578148383342215e-05,
+      "loss": 4.8604,
+      "step": 1304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.563885303616252e-05,
+      "loss": 4.3121,
+      "step": 1305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.549625206358403e-05,
+      "loss": 4.5801,
+      "step": 1306
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.5353681211835e-05,
+      "loss": 4.4235,
+      "step": 1307
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.521114077700105e-05,
+      "loss": 4.1041,
+      "step": 1308
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.506863105510476e-05,
+      "loss": 4.0871,
+      "step": 1309
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.492615234210483e-05,
+      "loss": 3.2724,
+      "step": 1310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.478370493389563e-05,
+      "loss": 4.1692,
+      "step": 1311
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.464128912630648e-05,
+      "loss": 5.7239,
+      "step": 1312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.449890521510108e-05,
+      "loss": 4.8659,
+      "step": 1313
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.435655349597689e-05,
+      "loss": 4.6029,
+      "step": 1314
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.42142342645646e-05,
+      "loss": 5.9711,
+      "step": 1315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.407194781642726e-05,
+      "loss": 5.5802,
+      "step": 1316
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.392969444705999e-05,
+      "loss": 4.506,
+      "step": 1317
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.378747445188907e-05,
+      "loss": 5.3925,
+      "step": 1318
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.364528812627164e-05,
+      "loss": 5.8907,
+      "step": 1319
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.350313576549476e-05,
+      "loss": 5.5348,
+      "step": 1320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.336101766477506e-05,
+      "loss": 4.5874,
+      "step": 1321
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.32189341192579e-05,
+      "loss": 5.1374,
+      "step": 1322
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.30768854240171e-05,
+      "loss": 4.9395,
+      "step": 1323
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.293487187405386e-05,
+      "loss": 4.9261,
+      "step": 1324
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.279289376429653e-05,
+      "loss": 4.6351,
+      "step": 1325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.26509513895998e-05,
+      "loss": 4.2997,
+      "step": 1326
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.25090450447442e-05,
+      "loss": 3.9342,
+      "step": 1327
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.236717502443535e-05,
+      "loss": 5.4713,
+      "step": 1328
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.222534162330355e-05,
+      "loss": 4.3734,
+      "step": 1329
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.208354513590288e-05,
+      "loss": 5.4414,
+      "step": 1330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.194178585671101e-05,
+      "loss": 4.8691,
+      "step": 1331
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.180006408012811e-05,
+      "loss": 5.2395,
+      "step": 1332
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.165838010047659e-05,
+      "loss": 5.0522,
+      "step": 1333
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.151673421200026e-05,
+      "loss": 5.4484,
+      "step": 1334
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.137512670886397e-05,
+      "loss": 5.9874,
+      "step": 1335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.123355788515269e-05,
+      "loss": 4.7616,
+      "step": 1336
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.10920280348712e-05,
+      "loss": 4.7961,
+      "step": 1337
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.095053745194324e-05,
+      "loss": 6.4074,
+      "step": 1338
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.080908643021107e-05,
+      "loss": 6.0865,
+      "step": 1339
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.066767526343476e-05,
+      "loss": 5.3222,
+      "step": 1340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.052630424529165e-05,
+      "loss": 4.2057,
+      "step": 1341
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.038497366937562e-05,
+      "loss": 4.1427,
+      "step": 1342
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.024368382919665e-05,
+      "loss": 4.1618,
+      "step": 1343
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.010243501818003e-05,
+      "loss": 5.2721,
+      "step": 1344
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.996122752966595e-05,
+      "loss": 4.7679,
+      "step": 1345
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.982006165690866e-05,
+      "loss": 4.2132,
+      "step": 1346
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.967893769307608e-05,
+      "loss": 6.7272,
+      "step": 1347
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.953785593124909e-05,
+      "loss": 4.3649,
+      "step": 1348
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.939681666442089e-05,
+      "loss": 5.5524,
+      "step": 1349
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.925582018549641e-05,
+      "loss": 5.5042,
+      "step": 1350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.911486678729179e-05,
+      "loss": 4.7637,
+      "step": 1351
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.897395676253362e-05,
+      "loss": 4.3316,
+      "step": 1352
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.883309040385851e-05,
+      "loss": 3.9166,
+      "step": 1353
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.869226800381224e-05,
+      "loss": 4.1233,
+      "step": 1354
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.855148985484946e-05,
+      "loss": 5.45,
+      "step": 1355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.841075624933285e-05,
+      "loss": 4.8906,
+      "step": 1356
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.827006747953255e-05,
+      "loss": 5.4032,
+      "step": 1357
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.812942383762562e-05,
+      "loss": 5.668,
+      "step": 1358
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.798882561569547e-05,
+      "loss": 3.5444,
+      "step": 1359
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.7848273105731e-05,
+      "loss": 5.6852,
+      "step": 1360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.77077665996264e-05,
+      "loss": 5.6541,
+      "step": 1361
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.756730638918011e-05,
+      "loss": 5.0972,
+      "step": 1362
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.74268927660946e-05,
+      "loss": 4.899,
+      "step": 1363
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.728652602197551e-05,
+      "loss": 6.2645,
+      "step": 1364
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.714620644833111e-05,
+      "loss": 4.7932,
+      "step": 1365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.700593433657172e-05,
+      "loss": 5.536,
+      "step": 1366
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.686570997800913e-05,
+      "loss": 4.9743,
+      "step": 1367
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.672553366385591e-05,
+      "loss": 5.5102,
+      "step": 1368
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.658540568522487e-05,
+      "loss": 4.9791,
+      "step": 1369
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.644532633312842e-05,
+      "loss": 3.7203,
+      "step": 1370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.630529589847805e-05,
+      "loss": 4.0206,
+      "step": 1371
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.616531467208362e-05,
+      "loss": 5.1073,
+      "step": 1372
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.602538294465273e-05,
+      "loss": 6.4864,
+      "step": 1373
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.588550100679033e-05,
+      "loss": 4.7621,
+      "step": 1374
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.57456691489978e-05,
+      "loss": 4.9378,
+      "step": 1375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.56058876616726e-05,
+      "loss": 4.7526,
+      "step": 1376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.546615683510767e-05,
+      "loss": 3.7859,
+      "step": 1377
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.532647695949053e-05,
+      "loss": 5.5153,
+      "step": 1378
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.51868483249031e-05,
+      "loss": 5.4043,
+      "step": 1379
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.504727122132078e-05,
+      "loss": 4.9669,
+      "step": 1380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.490774593861191e-05,
+      "loss": 4.7376,
+      "step": 1381
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.476827276653734e-05,
+      "loss": 5.4246,
+      "step": 1382
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.462885199474956e-05,
+      "loss": 5.6152,
+      "step": 1383
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.448948391279234e-05,
+      "loss": 3.9914,
+      "step": 1384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.435016881009995e-05,
+      "loss": 4.5247,
+      "step": 1385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.421090697599666e-05,
+      "loss": 4.8735,
+      "step": 1386
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.407169869969619e-05,
+      "loss": 5.3069,
+      "step": 1387
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.393254427030096e-05,
+      "loss": 4.8929,
+      "step": 1388
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.37934439768015e-05,
+      "loss": 4.7617,
+      "step": 1389
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.365439810807608e-05,
+      "loss": 5.1977,
+      "step": 1390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.351540695288977e-05,
+      "loss": 5.7407,
+      "step": 1391
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.337647079989412e-05,
+      "loss": 5.4025,
+      "step": 1392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.32375899376264e-05,
+      "loss": 3.7166,
+      "step": 1393
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.309876465450906e-05,
+      "loss": 4.224,
+      "step": 1394
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.295999523884921e-05,
+      "loss": 5.3467,
+      "step": 1395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.282128197883785e-05,
+      "loss": 3.6798,
+      "step": 1396
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.268262516254934e-05,
+      "loss": 3.6066,
+      "step": 1397
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.254402507794092e-05,
+      "loss": 6.1143,
+      "step": 1398
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.240548201285186e-05,
+      "loss": 4.5202,
+      "step": 1399
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.226699625500317e-05,
+      "loss": 5.1016,
+      "step": 1400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.212856809199675e-05,
+      "loss": 4.3679,
+      "step": 1401
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.19901978113149e-05,
+      "loss": 4.4327,
+      "step": 1402
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.185188570031978e-05,
+      "loss": 4.0243,
+      "step": 1403
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.17136320462527e-05,
+      "loss": 3.7914,
+      "step": 1404
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.157543713623352e-05,
+      "loss": 6.5484,
+      "step": 1405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.143730125726019e-05,
+      "loss": 4.9792,
+      "step": 1406
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.129922469620798e-05,
+      "loss": 4.7862,
+      "step": 1407
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.116120773982905e-05,
+      "loss": 4.5555,
+      "step": 1408
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.102325067475171e-05,
+      "loss": 4.0671,
+      "step": 1409
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.088535378747991e-05,
+      "loss": 5.3278,
+      "step": 1410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.074751736439269e-05,
+      "loss": 4.8334,
+      "step": 1411
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.060974169174346e-05,
+      "loss": 4.456,
+      "step": 1412
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.047202705565942e-05,
+      "loss": 5.5928,
+      "step": 1413
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.033437374214111e-05,
+      "loss": 4.259,
+      "step": 1414
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.019678203706163e-05,
+      "loss": 5.1696,
+      "step": 1415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.005925222616624e-05,
+      "loss": 5.2212,
+      "step": 1416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.992178459507153e-05,
+      "loss": 4.9839,
+      "step": 1417
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.9784379429265e-05,
+      "loss": 4.8708,
+      "step": 1418
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.964703701410456e-05,
+      "loss": 4.8921,
+      "step": 1419
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.950975763481757e-05,
+      "loss": 4.6446,
+      "step": 1420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.937254157650069e-05,
+      "loss": 3.5035,
+      "step": 1421
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.923538912411894e-05,
+      "loss": 3.7247,
+      "step": 1422
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.909830056250527e-05,
+      "loss": 5.9275,
+      "step": 1423
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.896127617636001e-05,
+      "loss": 4.3838,
+      "step": 1424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.882431625025015e-05,
+      "loss": 4.7743,
+      "step": 1425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.86874210686088e-05,
+      "loss": 5.6698,
+      "step": 1426
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.855059091573472e-05,
+      "loss": 4.0371,
+      "step": 1427
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.841382607579148e-05,
+      "loss": 6.1539,
+      "step": 1428
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.827712683280707e-05,
+      "loss": 5.2565,
+      "step": 1429
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.81404934706733e-05,
+      "loss": 5.2091,
+      "step": 1430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.800392627314505e-05,
+      "loss": 5.6129,
+      "step": 1431
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.786742552383994e-05,
+      "loss": 5.7058,
+      "step": 1432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.773099150623741e-05,
+      "loss": 5.9128,
+      "step": 1433
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.759462450367846e-05,
+      "loss": 4.506,
+      "step": 1434
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.745832479936491e-05,
+      "loss": 4.4943,
+      "step": 1435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.732209267635874e-05,
+      "loss": 4.9614,
+      "step": 1436
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.718592841758161e-05,
+      "loss": 4.5175,
+      "step": 1437
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.704983230581425e-05,
+      "loss": 5.4299,
+      "step": 1438
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.691380462369588e-05,
+      "loss": 6.577,
+      "step": 1439
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.677784565372358e-05,
+      "loss": 5.0158,
+      "step": 1440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.664195567825173e-05,
+      "loss": 7.7583,
+      "step": 1441
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.650613497949143e-05,
+      "loss": 5.5448,
+      "step": 1442
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.637038383950997e-05,
+      "loss": 4.3421,
+      "step": 1443
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.62347025402301e-05,
+      "loss": 4.5886,
+      "step": 1444
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.609909136342955e-05,
+      "loss": 5.296,
+      "step": 1445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.596355059074043e-05,
+      "loss": 4.5241,
+      "step": 1446
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.582808050364864e-05,
+      "loss": 5.3466,
+      "step": 1447
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.56926813834933e-05,
+      "loss": 5.3449,
+      "step": 1448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.555735351146608e-05,
+      "loss": 5.0487,
+      "step": 1449
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.542209716861073e-05,
+      "loss": 4.3978,
+      "step": 1450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.528691263582254e-05,
+      "loss": 5.0966,
+      "step": 1451
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.51518001938475e-05,
+      "loss": 3.9492,
+      "step": 1452
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.501676012328201e-05,
+      "loss": 3.7915,
+      "step": 1453
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.488179270457209e-05,
+      "loss": 4.6875,
+      "step": 1454
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.474689821801295e-05,
+      "loss": 3.9395,
+      "step": 1455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.461207694374827e-05,
+      "loss": 5.3686,
+      "step": 1456
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.447732916176975e-05,
+      "loss": 4.1966,
+      "step": 1457
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.43426551519164e-05,
+      "loss": 5.5098,
+      "step": 1458
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.420805519387412e-05,
+      "loss": 3.378,
+      "step": 1459
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.40735295671749e-05,
+      "loss": 4.3908,
+      "step": 1460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.393907855119646e-05,
+      "loss": 5.2041,
+      "step": 1461
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.380470242516147e-05,
+      "loss": 5.2012,
+      "step": 1462
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.367040146813721e-05,
+      "loss": 5.3592,
+      "step": 1463
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.353617595903472e-05,
+      "loss": 6.2132,
+      "step": 1464
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.340202617660842e-05,
+      "loss": 5.2942,
+      "step": 1465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.326795239945543e-05,
+      "loss": 6.7458,
+      "step": 1466
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.313395490601512e-05,
+      "loss": 6.117,
+      "step": 1467
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.30000339745683e-05,
+      "loss": 3.8988,
+      "step": 1468
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.286618988323688e-05,
+      "loss": 4.5317,
+      "step": 1469
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.273242290998311e-05,
+      "loss": 4.8566,
+      "step": 1470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.259873333260917e-05,
+      "loss": 5.4367,
+      "step": 1471
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.246512142875642e-05,
+      "loss": 3.7607,
+      "step": 1472
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.233158747590502e-05,
+      "loss": 5.1812,
+      "step": 1473
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.219813175137305e-05,
+      "loss": 3.9772,
+      "step": 1474
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.206475453231644e-05,
+      "loss": 4.8201,
+      "step": 1475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.193145609572777e-05,
+      "loss": 5.5693,
+      "step": 1476
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.179823671843621e-05,
+      "loss": 3.1308,
+      "step": 1477
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.16650966771066e-05,
+      "loss": 3.7313,
+      "step": 1478
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.153203624823917e-05,
+      "loss": 4.8219,
+      "step": 1479
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.139905570816867e-05,
+      "loss": 4.2721,
+      "step": 1480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.126615533306404e-05,
+      "loss": 6.0798,
+      "step": 1481
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.113333539892764e-05,
+      "loss": 5.5463,
+      "step": 1482
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.100059618159493e-05,
+      "loss": 5.4084,
+      "step": 1483
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.086793795673357e-05,
+      "loss": 4.9089,
+      "step": 1484
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.073536099984314e-05,
+      "loss": 3.7496,
+      "step": 1485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.0602865586254345e-05,
+      "loss": 6.45,
+      "step": 1486
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.047045199112864e-05,
+      "loss": 6.2711,
+      "step": 1487
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.0338120489457486e-05,
+      "loss": 4.4672,
+      "step": 1488
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.020587135606193e-05,
+      "loss": 4.0681,
+      "step": 1489
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 6.007370486559185e-05,
+      "loss": 4.0094,
+      "step": 1490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.9941621292525606e-05,
+      "loss": 3.3563,
+      "step": 1491
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.9809620911169325e-05,
+      "loss": 5.3856,
+      "step": 1492
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.967770399565638e-05,
+      "loss": 6.0677,
+      "step": 1493
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.9545870819946704e-05,
+      "loss": 7.3225,
+      "step": 1494
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.941412165782645e-05,
+      "loss": 4.4001,
+      "step": 1495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.928245678290723e-05,
+      "loss": 5.5827,
+      "step": 1496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.915087646862562e-05,
+      "loss": 4.9228,
+      "step": 1497
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.901938098824255e-05,
+      "loss": 4.7606,
+      "step": 1498
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.888797061484288e-05,
+      "loss": 3.8311,
+      "step": 1499
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.875664562133462e-05,
+      "loss": 5.5077,
+      "step": 1500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.86254062804485e-05,
+      "loss": 5.2451,
+      "step": 1501
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.849425286473737e-05,
+      "loss": 5.1935,
+      "step": 1502
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.836318564657561e-05,
+      "loss": 5.2688,
+      "step": 1503
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.823220489815867e-05,
+      "loss": 6.9505,
+      "step": 1504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.810131089150228e-05,
+      "loss": 4.2627,
+      "step": 1505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.79705038984421e-05,
+      "loss": 4.7349,
+      "step": 1506
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.7839784190633226e-05,
+      "loss": 5.7012,
+      "step": 1507
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.770915203954923e-05,
+      "loss": 5.4551,
+      "step": 1508
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.757860771648201e-05,
+      "loss": 4.7902,
+      "step": 1509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.744815149254102e-05,
+      "loss": 4.8647,
+      "step": 1510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.7317783638652775e-05,
+      "loss": 4.4612,
+      "step": 1511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.718750442556028e-05,
+      "loss": 5.8965,
+      "step": 1512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.7057314123822344e-05,
+      "loss": 3.1624,
+      "step": 1513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.6927213003813215e-05,
+      "loss": 5.0094,
+      "step": 1514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.679720133572206e-05,
+      "loss": 4.4758,
+      "step": 1515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.6667279389551985e-05,
+      "loss": 5.8778,
+      "step": 1516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.653744743511999e-05,
+      "loss": 5.7196,
+      "step": 1517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.640770574205614e-05,
+      "loss": 4.7232,
+      "step": 1518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.627805457980298e-05,
+      "loss": 3.4582,
+      "step": 1519
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.6148494217615134e-05,
+      "loss": 4.6131,
+      "step": 1520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.601902492455855e-05,
+      "loss": 4.7631,
+      "step": 1521
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.58896469695101e-05,
+      "loss": 5.1062,
+      "step": 1522
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.576036062115709e-05,
+      "loss": 4.0656,
+      "step": 1523
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.563116614799637e-05,
+      "loss": 5.9401,
+      "step": 1524
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.550206381833409e-05,
+      "loss": 3.223,
+      "step": 1525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.5373053900285067e-05,
+      "loss": 4.662,
+      "step": 1526
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.524413666177215e-05,
+      "loss": 3.9129,
+      "step": 1527
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.5115312370525764e-05,
+      "loss": 4.8251,
+      "step": 1528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.498658129408323e-05,
+      "loss": 4.8542,
+      "step": 1529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.485794369978828e-05,
+      "loss": 4.4469,
+      "step": 1530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.472939985479071e-05,
+      "loss": 3.6999,
+      "step": 1531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.4600950026045326e-05,
+      "loss": 4.0614,
+      "step": 1532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.447259448031188e-05,
+      "loss": 5.3122,
+      "step": 1533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.434433348415424e-05,
+      "loss": 5.2486,
+      "step": 1534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.421616730394e-05,
+      "loss": 5.4506,
+      "step": 1535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.40880962058397e-05,
+      "loss": 4.7842,
+      "step": 1536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.3960120455826545e-05,
+      "loss": 4.9233,
+      "step": 1537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.383224031967564e-05,
+      "loss": 4.3351,
+      "step": 1538
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.370445606296367e-05,
+      "loss": 4.8061,
+      "step": 1539
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.357676795106799e-05,
+      "loss": 5.5001,
+      "step": 1540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.3449176249166435e-05,
+      "loss": 5.2056,
+      "step": 1541
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.3321681222236566e-05,
+      "loss": 5.9456,
+      "step": 1542
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.319428313505522e-05,
+      "loss": 5.7687,
+      "step": 1543
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.306698225219781e-05,
+      "loss": 5.5028,
+      "step": 1544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.293977883803797e-05,
+      "loss": 4.6781,
+      "step": 1545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.281267315674685e-05,
+      "loss": 4.1077,
+      "step": 1546
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.26856654722928e-05,
+      "loss": 4.156,
+      "step": 1547
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.255875604844041e-05,
+      "loss": 5.2687,
+      "step": 1548
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.243194514875033e-05,
+      "loss": 6.1441,
+      "step": 1549
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.2305233036578615e-05,
+      "loss": 3.5804,
+      "step": 1550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.217861997507618e-05,
+      "loss": 5.9198,
+      "step": 1551
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.205210622718811e-05,
+      "loss": 3.7127,
+      "step": 1552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.1925692055653366e-05,
+      "loss": 5.958,
+      "step": 1553
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.179937772300401e-05,
+      "loss": 5.1324,
+      "step": 1554
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.167316349156495e-05,
+      "loss": 4.6707,
+      "step": 1555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.1547049623452936e-05,
+      "loss": 4.368,
+      "step": 1556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.142103638057652e-05,
+      "loss": 5.2238,
+      "step": 1557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.129512402463514e-05,
+      "loss": 4.7552,
+      "step": 1558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.116931281711885e-05,
+      "loss": 6.2025,
+      "step": 1559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 5.104360301930744e-05,
+      "loss": 4.8607,
+      "step": 1560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.091799489227027e-05,
+      "loss": 5.428,
+      "step": 1561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.0792488696865436e-05,
+      "loss": 4.4344,
+      "step": 1562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.066708469373957e-05,
+      "loss": 4.3613,
+      "step": 1563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.054178314332676e-05,
+      "loss": 5.0527,
+      "step": 1564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.0416584305848524e-05,
+      "loss": 7.0092,
+      "step": 1565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.029148844131303e-05,
+      "loss": 3.3616,
+      "step": 1566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.016649580951463e-05,
+      "loss": 3.4385,
+      "step": 1567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 5.0041606670033136e-05,
+      "loss": 4.4991,
+      "step": 1568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9916821282233615e-05,
+      "loss": 4.0545,
+      "step": 1569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9792139905265575e-05,
+      "loss": 4.2878,
+      "step": 1570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.966756279806255e-05,
+      "loss": 4.5769,
+      "step": 1571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9543090219341495e-05,
+      "loss": 5.232,
+      "step": 1572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.941872242760232e-05,
+      "loss": 5.0997,
+      "step": 1573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.929445968112728e-05,
+      "loss": 4.0502,
+      "step": 1574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.917030223798057e-05,
+      "loss": 5.3888,
+      "step": 1575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9046250356007516e-05,
+      "loss": 3.7761,
+      "step": 1576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8922304292834376e-05,
+      "loss": 5.6465,
+      "step": 1577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8798464305867594e-05,
+      "loss": 4.0709,
+      "step": 1578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.867473065229332e-05,
+      "loss": 4.8834,
+      "step": 1579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.855110358907689e-05,
+      "loss": 5.0408,
+      "step": 1580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.842758337296225e-05,
+      "loss": 4.0681,
+      "step": 1581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.830417026047147e-05,
+      "loss": 5.8733,
+      "step": 1582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.8180864507904224e-05,
+      "loss": 6.3516,
+      "step": 1583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.805766637133711e-05,
+      "loss": 4.924,
+      "step": 1584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.793457610662334e-05,
+      "loss": 6.0122,
+      "step": 1585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.781159396939209e-05,
+      "loss": 4.1038,
+      "step": 1586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.768872021504794e-05,
+      "loss": 4.868,
+      "step": 1587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.756595509877042e-05,
+      "loss": 5.258,
+      "step": 1588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.744329887551343e-05,
+      "loss": 3.3688,
+      "step": 1589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7320751800004725e-05,
+      "loss": 6.137,
+      "step": 1590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7198314126745426e-05,
+      "loss": 5.8388,
+      "step": 1591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.707598611000934e-05,
+      "loss": 5.2071,
+      "step": 1592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.695376800384266e-05,
+      "loss": 5.8804,
+      "step": 1593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.683166006206327e-05,
+      "loss": 5.3172,
+      "step": 1594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6709662538260267e-05,
+      "loss": 5.8974,
+      "step": 1595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.658777568579347e-05,
+      "loss": 4.9335,
+      "step": 1596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.646599975779281e-05,
+      "loss": 5.8098,
+      "step": 1597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.6344335007157935e-05,
+      "loss": 2.864,
+      "step": 1598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.622278168655748e-05,
+      "loss": 4.1246,
+      "step": 1599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.610134004842875e-05,
+      "loss": 4.8173,
+      "step": 1600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.598001034497712e-05,
+      "loss": 4.1158,
+      "step": 1601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5858792828175467e-05,
+      "loss": 4.9128,
+      "step": 1602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5737687749763714e-05,
+      "loss": 5.473,
+      "step": 1603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.561669536124823e-05,
+      "loss": 5.2666,
+      "step": 1604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.549581591390142e-05,
+      "loss": 4.9472,
+      "step": 1605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5375049658761104e-05,
+      "loss": 5.5872,
+      "step": 1606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.525439684662998e-05,
+      "loss": 4.6561,
+      "step": 1607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.51338577280752e-05,
+      "loss": 3.6993,
+      "step": 1608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.5013432553427834e-05,
+      "loss": 5.711,
+      "step": 1609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.489312157278227e-05,
+      "loss": 5.9926,
+      "step": 1610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.477292503599574e-05,
+      "loss": 4.625,
+      "step": 1611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.4652843192687844e-05,
+      "loss": 5.2421,
+      "step": 1612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.4532876292239936e-05,
+      "loss": 5.9339,
+      "step": 1613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.441302458379475e-05,
+      "loss": 5.6844,
+      "step": 1614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.4293288316255653e-05,
+      "loss": 3.2956,
+      "step": 1615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.4173667738286375e-05,
+      "loss": 4.3837,
+      "step": 1616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.405416309831037e-05,
+      "loss": 5.4053,
+      "step": 1617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.3934774644510324e-05,
+      "loss": 5.1529,
+      "step": 1618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.381550262482759e-05,
+      "loss": 6.3204,
+      "step": 1619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.3696347286961744e-05,
+      "loss": 5.7969,
+      "step": 1620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.357730887837005e-05,
+      "loss": 3.968,
+      "step": 1621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.345838764626694e-05,
+      "loss": 5.6015,
+      "step": 1622
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.333958383762344e-05,
+      "loss": 6.9745,
+      "step": 1623
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.3220897699166775e-05,
+      "loss": 6.25,
+      "step": 1624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.310232947737979e-05,
+      "loss": 5.2227,
+      "step": 1625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.298387941850043e-05,
+      "loss": 5.5634,
+      "step": 1626
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.286554776852125e-05,
+      "loss": 5.1307,
+      "step": 1627
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.2747334773188916e-05,
+      "loss": 4.3987,
+      "step": 1628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.262924067800364e-05,
+      "loss": 4.2386,
+      "step": 1629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.251126572821876e-05,
+      "loss": 4.6875,
+      "step": 1630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.239341016884009e-05,
+      "loss": 3.6091,
+      "step": 1631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.2275674244625576e-05,
+      "loss": 5.5601,
+      "step": 1632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.215805820008467e-05,
+      "loss": 5.3921,
+      "step": 1633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.204056227947791e-05,
+      "loss": 4.9879,
+      "step": 1634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.192318672681631e-05,
+      "loss": 4.6314,
+      "step": 1635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.180593178586094e-05,
+      "loss": 4.4372,
+      "step": 1636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.168879770012237e-05,
+      "loss": 6.6243,
+      "step": 1637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.157178471286025e-05,
+      "loss": 4.8991,
+      "step": 1638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.14548930670826e-05,
+      "loss": 4.2745,
+      "step": 1639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.133812300554555e-05,
+      "loss": 4.1917,
+      "step": 1640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.12214747707527e-05,
+      "loss": 6.0976,
+      "step": 1641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.110494860495465e-05,
+      "loss": 3.1873,
+      "step": 1642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.0988544750148495e-05,
+      "loss": 5.466,
+      "step": 1643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.0872263448077296e-05,
+      "loss": 4.0799,
+      "step": 1644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.075610494022964e-05,
+      "loss": 4.149,
+      "step": 1645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.064006946783909e-05,
+      "loss": 5.4425,
+      "step": 1646
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.0524157271883634e-05,
+      "loss": 4.0427,
+      "step": 1647
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.04083685930853e-05,
+      "loss": 3.4482,
+      "step": 1648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.0292703671909607e-05,
+      "loss": 4.0471,
+      "step": 1649
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.017716274856502e-05,
+      "loss": 5.0375,
+      "step": 1650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.006174606300255e-05,
+      "loss": 4.5553,
+      "step": 1651
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.994645385491513e-05,
+      "loss": 4.3989,
+      "step": 1652
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.983128636373723e-05,
+      "loss": 3.2933,
+      "step": 1653
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.971624382864432e-05,
+      "loss": 4.5164,
+      "step": 1654
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.960132648855226e-05,
+      "loss": 5.6185,
+      "step": 1655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.948653458211704e-05,
+      "loss": 5.7126,
+      "step": 1656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.93718683477341e-05,
+      "loss": 3.5738,
+      "step": 1657
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.9257328023537884e-05,
+      "loss": 4.102,
+      "step": 1658
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.914291384740138e-05,
+      "loss": 3.7296,
+      "step": 1659
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.902862605693557e-05,
+      "loss": 4.1262,
+      "step": 1660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.8914464889488945e-05,
+      "loss": 4.0301,
+      "step": 1661
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.880043058214712e-05,
+      "loss": 5.205,
+      "step": 1662
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.868652337173211e-05,
+      "loss": 5.2681,
+      "step": 1663
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.857274349480208e-05,
+      "loss": 4.8541,
+      "step": 1664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.845909118765073e-05,
+      "loss": 5.1282,
+      "step": 1665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.834556668630684e-05,
+      "loss": 5.769,
+      "step": 1666
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.823217022653376e-05,
+      "loss": 4.4028,
+      "step": 1667
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.81189020438289e-05,
+      "loss": 5.0993,
+      "step": 1668
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.800576237342335e-05,
+      "loss": 5.6765,
+      "step": 1669
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.789275145028116e-05,
+      "loss": 4.2736,
+      "step": 1670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.777986950909917e-05,
+      "loss": 5.8832,
+      "step": 1671
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.7667116784306235e-05,
+      "loss": 5.9556,
+      "step": 1672
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.755449351006294e-05,
+      "loss": 3.9721,
+      "step": 1673
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.744199992026098e-05,
+      "loss": 6.1241,
+      "step": 1674
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.732963624852275e-05,
+      "loss": 5.5369,
+      "step": 1675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.7217402728200824e-05,
+      "loss": 6.1986,
+      "step": 1676
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.710529959237753e-05,
+      "loss": 5.024,
+      "step": 1677
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.699332707386429e-05,
+      "loss": 4.9638,
+      "step": 1678
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.688148540520141e-05,
+      "loss": 6.2853,
+      "step": 1679
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.6769774818657374e-05,
+      "loss": 4.6099,
+      "step": 1680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.665819554622847e-05,
+      "loss": 5.5092,
+      "step": 1681
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.654674781963826e-05,
+      "loss": 5.5631,
+      "step": 1682
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.643543187033712e-05,
+      "loss": 4.1108,
+      "step": 1683
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.6324247929501765e-05,
+      "loss": 5.595,
+      "step": 1684
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.62131962280348e-05,
+      "loss": 5.6932,
+      "step": 1685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.6102276996564054e-05,
+      "loss": 5.6078,
+      "step": 1686
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.599149046544241e-05,
+      "loss": 3.6451,
+      "step": 1687
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.5880836864747105e-05,
+      "loss": 4.159,
+      "step": 1688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.577031642427928e-05,
+      "loss": 4.277,
+      "step": 1689
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.5659929373563586e-05,
+      "loss": 3.6792,
+      "step": 1690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.5549675941847614e-05,
+      "loss": 4.7623,
+      "step": 1691
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.54395563581015e-05,
+      "loss": 5.1786,
+      "step": 1692
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.53295708510174e-05,
+      "loss": 6.1749,
+      "step": 1693
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.521971964900895e-05,
+      "loss": 3.3187,
+      "step": 1694
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.5110002980210975e-05,
+      "loss": 5.6126,
+      "step": 1695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.500042107247884e-05,
+      "loss": 4.2123,
+      "step": 1696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.4890974153388076e-05,
+      "loss": 4.167,
+      "step": 1697
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.478166245023385e-05,
+      "loss": 4.7567,
+      "step": 1698
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.467248619003054e-05,
+      "loss": 4.0213,
+      "step": 1699
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.456344559951122e-05,
+      "loss": 6.0049,
+      "step": 1700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.4454540905127274e-05,
+      "loss": 6.1917,
+      "step": 1701
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.434577233304772e-05,
+      "loss": 4.5018,
+      "step": 1702
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.4237140109159014e-05,
+      "loss": 5.5831,
+      "step": 1703
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.412864445906442e-05,
+      "loss": 3.7823,
+      "step": 1704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.4020285608083566e-05,
+      "loss": 4.9256,
+      "step": 1705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.391206378125189e-05,
+      "loss": 5.5552,
+      "step": 1706
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.3803979203320446e-05,
+      "loss": 4.305,
+      "step": 1707
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.36960320987551e-05,
+      "loss": 5.952,
+      "step": 1708
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.358822269173634e-05,
+      "loss": 4.1163,
+      "step": 1709
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.348055120615854e-05,
+      "loss": 4.6827,
+      "step": 1710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.337301786562974e-05,
+      "loss": 7.2957,
+      "step": 1711
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.3265622893471104e-05,
+      "loss": 4.4287,
+      "step": 1712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.315836651271642e-05,
+      "loss": 5.5463,
+      "step": 1713
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.3051248946111554e-05,
+      "loss": 4.8691,
+      "step": 1714
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.294427041611425e-05,
+      "loss": 3.6256,
+      "step": 1715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.2837431144893425e-05,
+      "loss": 5.9513,
+      "step": 1716
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.273073135432882e-05,
+      "loss": 4.9251,
+      "step": 1717
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.2624171266010405e-05,
+      "loss": 4.4213,
+      "step": 1718
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.251775110123814e-05,
+      "loss": 4.5091,
+      "step": 1719
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.241147108102135e-05,
+      "loss": 3.5291,
+      "step": 1720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.2305331426078364e-05,
+      "loss": 3.5106,
+      "step": 1721
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.2199332356835854e-05,
+      "loss": 4.6154,
+      "step": 1722
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.209347409342873e-05,
+      "loss": 3.1258,
+      "step": 1723
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.198775685569937e-05,
+      "loss": 5.7532,
+      "step": 1724
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.188218086319731e-05,
+      "loss": 4.2749,
+      "step": 1725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.177674633517868e-05,
+      "loss": 4.3017,
+      "step": 1726
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.167145349060591e-05,
+      "loss": 3.8278,
+      "step": 1727
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.1566302548147155e-05,
+      "loss": 5.7521,
+      "step": 1728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.146129372617589e-05,
+      "loss": 4.2083,
+      "step": 1729
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.135642724277036e-05,
+      "loss": 5.2714,
+      "step": 1730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.125170331571333e-05,
+      "loss": 4.6812,
+      "step": 1731
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.114712216249145e-05,
+      "loss": 5.8599,
+      "step": 1732
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.104268400029487e-05,
+      "loss": 3.5029,
+      "step": 1733
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.093838904601672e-05,
+      "loss": 4.1452,
+      "step": 1734
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.083423751625282e-05,
+      "loss": 4.0667,
+      "step": 1735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.073022962730105e-05,
+      "loss": 5.8295,
+      "step": 1736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.06263655951611e-05,
+      "loss": 4.3601,
+      "step": 1737
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.052264563553367e-05,
+      "loss": 4.528,
+      "step": 1738
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.0419069963820556e-05,
+      "loss": 4.9895,
+      "step": 1739
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.0315638795123725e-05,
+      "loss": 4.8056,
+      "step": 1740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.021235234424502e-05,
+      "loss": 5.1182,
+      "step": 1741
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.010921082568583e-05,
+      "loss": 4.9657,
+      "step": 1742
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.000621445364651e-05,
+      "loss": 5.2988,
+      "step": 1743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9903363442026e-05,
+      "loss": 6.0085,
+      "step": 1744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9800658004421366e-05,
+      "loss": 5.4803,
+      "step": 1745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9698098354127257e-05,
+      "loss": 5.6367,
+      "step": 1746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.959568470413572e-05,
+      "loss": 4.956,
+      "step": 1747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9493417267135514e-05,
+      "loss": 5.4248,
+      "step": 1748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.939129625551168e-05,
+      "loss": 6.1208,
+      "step": 1749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9289321881345254e-05,
+      "loss": 5.3616,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9187494356412736e-05,
+      "loss": 4.7495,
+      "step": 1751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9085813892185632e-05,
+      "loss": 4.4627,
+      "step": 1752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8984280699830068e-05,
+      "loss": 4.5524,
+      "step": 1753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8882894990206243e-05,
+      "loss": 3.7172,
+      "step": 1754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.878165697386812e-05,
+      "loss": 4.5049,
+      "step": 1755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.868056686106304e-05,
+      "loss": 3.1222,
+      "step": 1756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.857962486173098e-05,
+      "loss": 3.8222,
+      "step": 1757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.847883118550447e-05,
+      "loss": 4.5516,
+      "step": 1758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8378186041707977e-05,
+      "loss": 4.3736,
+      "step": 1759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8277689639357452e-05,
+      "loss": 5.0432,
+      "step": 1760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8177342187160048e-05,
+      "loss": 5.1772,
+      "step": 1761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.8077143893513447e-05,
+      "loss": 3.5626,
+      "step": 1762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7977094966505624e-05,
+      "loss": 2.2183,
+      "step": 1763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7877195613914475e-05,
+      "loss": 4.1043,
+      "step": 1764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7777446043207058e-05,
+      "loss": 4.8657,
+      "step": 1765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.767784646153949e-05,
+      "loss": 5.0821,
+      "step": 1766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7578397075756403e-05,
+      "loss": 4.6695,
+      "step": 1767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.747909809239049e-05,
+      "loss": 4.5632,
+      "step": 1768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7379949717662013e-05,
+      "loss": 5.3534,
+      "step": 1769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.728095215747859e-05,
+      "loss": 5.7769,
+      "step": 1770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.7182105617434518e-05,
+      "loss": 4.9148,
+      "step": 1771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.708341030281061e-05,
+      "loss": 5.6066,
+      "step": 1772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.698486641857343e-05,
+      "loss": 4.9809,
+      "step": 1773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6886474169375188e-05,
+      "loss": 4.0965,
+      "step": 1774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.678823375955314e-05,
+      "loss": 5.5814,
+      "step": 1775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.669014539312924e-05,
+      "loss": 4.0284,
+      "step": 1776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.659220927380961e-05,
+      "loss": 4.2508,
+      "step": 1777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6494425604984263e-05,
+      "loss": 2.8847,
+      "step": 1778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.639679458972657e-05,
+      "loss": 4.3559,
+      "step": 1779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6299316430792952e-05,
+      "loss": 5.1336,
+      "step": 1780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6201991330622265e-05,
+      "loss": 5.0795,
+      "step": 1781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6104819491335562e-05,
+      "loss": 4.1207,
+      "step": 1782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6007801114735632e-05,
+      "loss": 4.8409,
+      "step": 1783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.591093640230654e-05,
+      "loss": 4.6625,
+      "step": 1784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5814225555213156e-05,
+      "loss": 4.6359,
+      "step": 1785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5717668774300917e-05,
+      "loss": 6.0273,
+      "step": 1786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5621266260095224e-05,
+      "loss": 5.4293,
+      "step": 1787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5525018212801222e-05,
+      "loss": 4.1165,
+      "step": 1788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.542892483230308e-05,
+      "loss": 5.8879,
+      "step": 1789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5332986318163886e-05,
+      "loss": 4.1292,
+      "step": 1790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.52372028696251e-05,
+      "loss": 5.3274,
+      "step": 1791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.514157468560613e-05,
+      "loss": 4.3405,
+      "step": 1792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5046101964703893e-05,
+      "loss": 4.7678,
+      "step": 1793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4950784905192514e-05,
+      "loss": 5.6349,
+      "step": 1794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.485562370502279e-05,
+      "loss": 3.9829,
+      "step": 1795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4760618561821945e-05,
+      "loss": 4.5781,
+      "step": 1796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4665769672892947e-05,
+      "loss": 4.1321,
+      "step": 1797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4571077235214368e-05,
+      "loss": 4.3518,
+      "step": 1798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.447654144543986e-05,
+      "loss": 5.4984,
+      "step": 1799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4382162499897742e-05,
+      "loss": 4.3747,
+      "step": 1800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4287940594590564e-05,
+      "loss": 5.2658,
+      "step": 1801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4193875925194797e-05,
+      "loss": 4.6805,
+      "step": 1802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.409996868706036e-05,
+      "loss": 4.982,
+      "step": 1803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4006219075210203e-05,
+      "loss": 4.0818,
+      "step": 1804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.391262728433995e-05,
+      "loss": 5.2595,
+      "step": 1805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.381919350881745e-05,
+      "loss": 4.6005,
+      "step": 1806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3725917942682397e-05,
+      "loss": 4.275,
+      "step": 1807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3632800779645935e-05,
+      "loss": 3.3194,
+      "step": 1808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3539842213090202e-05,
+      "loss": 3.5953,
+      "step": 1809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.344704243606801e-05,
+      "loss": 5.16,
+      "step": 1810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3354401641302394e-05,
+      "loss": 4.5631,
+      "step": 1811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.32619200211862e-05,
+      "loss": 3.7543,
+      "step": 1812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.3169597767781737e-05,
+      "loss": 4.382,
+      "step": 1813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.307743507282034e-05,
+      "loss": 4.9816,
+      "step": 1814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2985432127701946e-05,
+      "loss": 4.0126,
+      "step": 1815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.289358912349481e-05,
+      "loss": 4.6121,
+      "step": 1816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2801906250934902e-05,
+      "loss": 5.458,
+      "step": 1817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2710383700425743e-05,
+      "loss": 3.8737,
+      "step": 1818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2619021662037854e-05,
+      "loss": 4.3943,
+      "step": 1819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2527820325508442e-05,
+      "loss": 5.2549,
+      "step": 1820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2436779880240943e-05,
+      "loss": 4.4907,
+      "step": 1821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2345900515304696e-05,
+      "loss": 4.7963,
+      "step": 1822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.225518241943446e-05,
+      "loss": 4.8078,
+      "step": 1823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2164625781030168e-05,
+      "loss": 5.1665,
+      "step": 1824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.2074230788156326e-05,
+      "loss": 3.3301,
+      "step": 1825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1983997628541808e-05,
+      "loss": 5.6025,
+      "step": 1826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1893926489579397e-05,
+      "loss": 4.4551,
+      "step": 1827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1804017558325406e-05,
+      "loss": 6.1391,
+      "step": 1828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1714271021499265e-05,
+      "loss": 4.6996,
+      "step": 1829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.162468706548314e-05,
+      "loss": 6.3854,
+      "step": 1830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1535265876321576e-05,
+      "loss": 5.613,
+      "step": 1831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1446007639721112e-05,
+      "loss": 3.852,
+      "step": 1832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1356912541049768e-05,
+      "loss": 4.3412,
+      "step": 1833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1267980765336893e-05,
+      "loss": 5.5092,
+      "step": 1834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1179212497272584e-05,
+      "loss": 5.4852,
+      "step": 1835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1090607921207396e-05,
+      "loss": 5.7496,
+      "step": 1836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.1002167221151914e-05,
+      "loss": 5.619,
+      "step": 1837
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.091389058077642e-05,
+      "loss": 4.4377,
+      "step": 1838
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0825778183410483e-05,
+      "loss": 3.8635,
+      "step": 1839
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.073783021204251e-05,
+      "loss": 4.6316,
+      "step": 1840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0650046849319516e-05,
+      "loss": 3.3127,
+      "step": 1841
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.056242827754663e-05,
+      "loss": 4.426,
+      "step": 1842
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0474974678686764e-05,
+      "loss": 5.3123,
+      "step": 1843
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0387686234360193e-05,
+      "loss": 4.4512,
+      "step": 1844
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.030056312584424e-05,
+      "loss": 5.3663,
+      "step": 1845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0213605534072833e-05,
+      "loss": 6.9047,
+      "step": 1846
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.012681363963621e-05,
+      "loss": 3.9307,
+      "step": 1847
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.0040187622780405e-05,
+      "loss": 5.2575,
+      "step": 1848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9953727663407042e-05,
+      "loss": 3.3696,
+      "step": 1849
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9867433941072855e-05,
+      "loss": 4.9946,
+      "step": 1850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9781306634989372e-05,
+      "loss": 4.2543,
+      "step": 1851
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.969534592402247e-05,
+      "loss": 4.6051,
+      "step": 1852
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9609551986692064e-05,
+      "loss": 5.0115,
+      "step": 1853
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.952392500117175e-05,
+      "loss": 4.0383,
+      "step": 1854
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9438465145288374e-05,
+      "loss": 4.337,
+      "step": 1855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9353172596521672e-05,
+      "loss": 4.8764,
+      "step": 1856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9268047532003963e-05,
+      "loss": 8.0361,
+      "step": 1857
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9183090128519732e-05,
+      "loss": 4.4439,
+      "step": 1858
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9098300562505266e-05,
+      "loss": 5.885,
+      "step": 1859
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.901367901004829e-05,
+      "loss": 5.0705,
+      "step": 1860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.8929225646887617e-05,
+      "loss": 5.4422,
+      "step": 1861
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.8844940648412767e-05,
+      "loss": 3.3901,
+      "step": 1862
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.8760824189663617e-05,
+      "loss": 4.5233,
+      "step": 1863
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.8676876445329973e-05,
+      "loss": 3.7609,
+      "step": 1864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.859309758975132e-05,
+      "loss": 4.4295,
+      "step": 1865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.850948779691639e-05,
+      "loss": 4.3773,
+      "step": 1866
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.8426047240462806e-05,
+      "loss": 6.3442,
+      "step": 1867
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.8342776093676715e-05,
+      "loss": 3.6569,
+      "step": 1868
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.8259674529492475e-05,
+      "loss": 5.9922,
+      "step": 1869
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.817674272049221e-05,
+      "loss": 4.6109,
+      "step": 1870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.809398083890559e-05,
+      "loss": 4.5278,
+      "step": 1871
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.801138905660924e-05,
+      "loss": 4.7826,
+      "step": 1872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.7928967545126675e-05,
+      "loss": 4.069,
+      "step": 1873
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.7846716475627713e-05,
+      "loss": 4.304,
+      "step": 1874
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.776463601892825e-05,
+      "loss": 4.5679,
+      "step": 1875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.7682726345489842e-05,
+      "loss": 5.8646,
+      "step": 1876
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.760098762541935e-05,
+      "loss": 5.8971,
+      "step": 1877
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.7519420028468647e-05,
+      "loss": 4.4435,
+      "step": 1878
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.7438023724034213e-05,
+      "loss": 5.2608,
+      "step": 1879
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.7356798881156734e-05,
+      "loss": 5.5128,
+      "step": 1880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.7275745668520883e-05,
+      "loss": 4.2629,
+      "step": 1881
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.719486425445487e-05,
+      "loss": 5.4742,
+      "step": 1882
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.711415480693014e-05,
+      "loss": 5.3501,
+      "step": 1883
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.703361749356098e-05,
+      "loss": 5.3519,
+      "step": 1884
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.69532524816042e-05,
+      "loss": 5.913,
+      "step": 1885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.687305993795879e-05,
+      "loss": 4.1534,
+      "step": 1886
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.6793040029165597e-05,
+      "loss": 5.3755,
+      "step": 1887
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.6713192921406827e-05,
+      "loss": 3.1755,
+      "step": 1888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.6633518780505965e-05,
+      "loss": 5.4903,
+      "step": 1889
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.6554017771927187e-05,
+      "loss": 6.1072,
+      "step": 1890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.647469006077518e-05,
+      "loss": 3.4047,
+      "step": 1891
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.639553581179468e-05,
+      "loss": 4.4568,
+      "step": 1892
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.631655518937022e-05,
+      "loss": 3.8945,
+      "step": 1893
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.623774835752574e-05,
+      "loss": 4.8321,
+      "step": 1894
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.6159115479924257e-05,
+      "loss": 5.0706,
+      "step": 1895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.6080656719867504e-05,
+      "loss": 3.3174,
+      "step": 1896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.600237224029565e-05,
+      "loss": 5.8968,
+      "step": 1897
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.592426220378691e-05,
+      "loss": 5.6086,
+      "step": 1898
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.584632677255723e-05,
+      "loss": 2.5947,
+      "step": 1899
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5768566108459913e-05,
+      "loss": 5.5297,
+      "step": 1900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5690980372985352e-05,
+      "loss": 5.1537,
+      "step": 1901
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5613569727260646e-05,
+      "loss": 6.7545,
+      "step": 1902
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5536334332049274e-05,
+      "loss": 4.7539,
+      "step": 1903
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5459274347750706e-05,
+      "loss": 5.1994,
+      "step": 1904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5382389934400198e-05,
+      "loss": 4.2053,
+      "step": 1905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.530568125166837e-05,
+      "loss": 4.9128,
+      "step": 1906
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5229148458860864e-05,
+      "loss": 3.9767,
+      "step": 1907
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.515279171491808e-05,
+      "loss": 5.4377,
+      "step": 1908
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5076611178414769e-05,
+      "loss": 5.332,
+      "step": 1909
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.5000607007559797e-05,
+      "loss": 4.5482,
+      "step": 1910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4924779360195662e-05,
+      "loss": 3.4208,
+      "step": 1911
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4849128393798351e-05,
+      "loss": 4.6318,
+      "step": 1912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.47736542654769e-05,
+      "loss": 3.9282,
+      "step": 1913
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4698357131973095e-05,
+      "loss": 4.7118,
+      "step": 1914
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.462323714966114e-05,
+      "loss": 5.6741,
+      "step": 1915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4548294474547352e-05,
+      "loss": 4.4112,
+      "step": 1916
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4473529262269814e-05,
+      "loss": 4.0998,
+      "step": 1917
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4398941668098077e-05,
+      "loss": 5.588,
+      "step": 1918
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4324531846932765e-05,
+      "loss": 3.7413,
+      "step": 1919
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4250299953305369e-05,
+      "loss": 4.4236,
+      "step": 1920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4176246141377858e-05,
+      "loss": 5.7991,
+      "step": 1921
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.4102370564942346e-05,
+      "loss": 3.7537,
+      "step": 1922
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.402867337742082e-05,
+      "loss": 3.3586,
+      "step": 1923
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3955154731864773e-05,
+      "loss": 4.8064,
+      "step": 1924
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3881814780954916e-05,
+      "loss": 4.9561,
+      "step": 1925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3808653677000894e-05,
+      "loss": 4.6645,
+      "step": 1926
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.3735671571940834e-05,
+      "loss": 4.0884,
+      "step": 1927
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3662868617341207e-05,
+      "loss": 5.1485,
+      "step": 1928
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3590244964396404e-05,
+      "loss": 4.8424,
+      "step": 1929
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3517800763928457e-05,
+      "loss": 3.3851,
+      "step": 1930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3445536166386707e-05,
+      "loss": 5.2136,
+      "step": 1931
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3373451321847518e-05,
+      "loss": 4.1184,
+      "step": 1932
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3301546380013941e-05,
+      "loss": 4.31,
+      "step": 1933
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3229821490215421e-05,
+      "loss": 4.724,
+      "step": 1934
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3158276801407431e-05,
+      "loss": 5.8546,
+      "step": 1935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3086912462171264e-05,
+      "loss": 4.6042,
+      "step": 1936
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3015728620713651e-05,
+      "loss": 4.93,
+      "step": 1937
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.294472542486651e-05,
+      "loss": 4.0118,
+      "step": 1938
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2873903022086486e-05,
+      "loss": 4.8937,
+      "step": 1939
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2803261559454904e-05,
+      "loss": 3.3115,
+      "step": 1940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2732801183677245e-05,
+      "loss": 4.576,
+      "step": 1941
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2662522041082925e-05,
+      "loss": 4.8809,
+      "step": 1942
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2592424277624947e-05,
+      "loss": 3.7892,
+      "step": 1943
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.25225080388797e-05,
+      "loss": 3.9746,
+      "step": 1944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2452773470046541e-05,
+      "loss": 4.5978,
+      "step": 1945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2383220715947597e-05,
+      "loss": 4.9512,
+      "step": 1946
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2313849921027277e-05,
+      "loss": 4.8203,
+      "step": 1947
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2244661229352283e-05,
+      "loss": 3.106,
+      "step": 1948
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2175654784611024e-05,
+      "loss": 5.1265,
+      "step": 1949
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.210683073011346e-05,
+      "loss": 4.0584,
+      "step": 1950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.2038189208790717e-05,
+      "loss": 4.6671,
+      "step": 1951
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1969730363194909e-05,
+      "loss": 5.4345,
+      "step": 1952
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1901454335498751e-05,
+      "loss": 3.7604,
+      "step": 1953
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1833361267495291e-05,
+      "loss": 4.3795,
+      "step": 1954
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1765451300597573e-05,
+      "loss": 5.0689,
+      "step": 1955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1697724575838465e-05,
+      "loss": 4.3988,
+      "step": 1956
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1630181233870229e-05,
+      "loss": 4.3237,
+      "step": 1957
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1562821414964309e-05,
+      "loss": 5.2297,
+      "step": 1958
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1495645259010968e-05,
+      "loss": 4.5607,
+      "step": 1959
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.14286529055191e-05,
+      "loss": 3.9856,
+      "step": 1960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1361844493615859e-05,
+      "loss": 3.2572,
+      "step": 1961
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1295220162046427e-05,
+      "loss": 5.6079,
+      "step": 1962
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1228780049173615e-05,
+      "loss": 4.5445,
+      "step": 1963
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.116252429297775e-05,
+      "loss": 6.2463,
+      "step": 1964
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.1096453031056264e-05,
+      "loss": 3.8266,
+      "step": 1965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.103056640062342e-05,
+      "loss": 5.0029,
+      "step": 1966
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0964864538510023e-05,
+      "loss": 4.9665,
+      "step": 1967
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0899347581163221e-05,
+      "loss": 6.0629,
+      "step": 1968
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0834015664646113e-05,
+      "loss": 3.5788,
+      "step": 1969
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0768868924637554e-05,
+      "loss": 4.3148,
+      "step": 1970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0703907496431741e-05,
+      "loss": 5.6798,
+      "step": 1971
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0639131514938138e-05,
+      "loss": 5.9043,
+      "step": 1972
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0574541114681047e-05,
+      "loss": 4.8213,
+      "step": 1973
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.051013642979929e-05,
+      "loss": 4.9091,
+      "step": 1974
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0445917594046072e-05,
+      "loss": 6.0368,
+      "step": 1975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0381884740788628e-05,
+      "loss": 3.828,
+      "step": 1976
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0318038003007923e-05,
+      "loss": 3.2175,
+      "step": 1977
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0254377513298463e-05,
+      "loss": 4.8269,
+      "step": 1978
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0190903403867846e-05,
+      "loss": 5.0557,
+      "step": 1979
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.012761580653674e-05,
+      "loss": 3.3731,
+      "step": 1980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0064514852738405e-05,
+      "loss": 4.8758,
+      "step": 1981
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.0001600673518462e-05,
+      "loss": 5.0746,
+      "step": 1982
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.938873399534687e-06,
+      "loss": 4.7923,
+      "step": 1983
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.876333161056672e-06,
+      "loss": 4.2104,
+      "step": 1984
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.813980087965623e-06,
+      "loss": 3.5032,
+      "step": 1985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.751814309754003e-06,
+      "loss": 4.0186,
+      "step": 1986
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689835955525305e-06,
+      "loss": 5.7008,
+      "step": 1987
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.628045153993837e-06,
+      "loss": 3.6164,
+      "step": 1988
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.56644203348438e-06,
+      "loss": 4.8254,
+      "step": 1989
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.505026721931897e-06,
+      "loss": 5.5035,
+      "step": 1990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.443799346881388e-06,
+      "loss": 4.4476,
+      "step": 1991
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.382760035487536e-06,
+      "loss": 5.9311,
+      "step": 1992
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.32190891451442e-06,
+      "loss": 5.5814,
+      "step": 1993
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.261246110335375e-06,
+      "loss": 4.7055,
+      "step": 1994
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.200771748932513e-06,
+      "loss": 5.5298,
+      "step": 1995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.140485955896738e-06,
+      "loss": 3.9145,
+      "step": 1996
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.08038885642729e-06,
+      "loss": 4.5976,
+      "step": 1997
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.020480575331481e-06,
+      "loss": 4.6587,
+      "step": 1998
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.960761237024573e-06,
+      "loss": 5.2935,
+      "step": 1999
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.901230965529383e-06,
+      "loss": 4.925,
+      "step": 2000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.841889884476117e-06,
+      "loss": 3.3641,
+      "step": 2001
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.782738117102085e-06,
+      "loss": 4.871,
+      "step": 2002
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.723775786251353e-06,
+      "loss": 3.2402,
+      "step": 2003
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.665003014374706e-06,
+      "loss": 3.6264,
+      "step": 2004
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.606419923529174e-06,
+      "loss": 5.3776,
+      "step": 2005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.548026635377848e-06,
+      "loss": 5.404,
+      "step": 2006
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.489823271189711e-06,
+      "loss": 4.4905,
+      "step": 2007
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.431809951839265e-06,
+      "loss": 4.9913,
+      "step": 2008
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.373986797806377e-06,
+      "loss": 4.7687,
+      "step": 2009
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.316353929175934e-06,
+      "loss": 4.7442,
+      "step": 2010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.258911465637675e-06,
+      "loss": 3.3546,
+      "step": 2011
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.20165952648595e-06,
+      "loss": 6.7167,
+      "step": 2012
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.144598230619383e-06,
+      "loss": 5.1568,
+      "step": 2013
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.08772769654068e-06,
+      "loss": 3.9749,
+      "step": 2014
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 8.031048042356392e-06,
+      "loss": 3.0926,
+      "step": 2015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.974559385776659e-06,
+      "loss": 4.7809,
+      "step": 2016
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.918261844114993e-06,
+      "loss": 4.8348,
+      "step": 2017
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.86215553428794e-06,
+      "loss": 4.5241,
+      "step": 2018
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.806240572814927e-06,
+      "loss": 4.2688,
+      "step": 2019
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.750517075818064e-06,
+      "loss": 5.1599,
+      "step": 2020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.694985159021761e-06,
+      "loss": 5.0288,
+      "step": 2021
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.639644937752565e-06,
+      "loss": 5.0567,
+      "step": 2022
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.584496526938934e-06,
+      "loss": 5.4037,
+      "step": 2023
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.529540041110983e-06,
+      "loss": 5.9251,
+      "step": 2024
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.474775594400252e-06,
+      "loss": 5.985,
+      "step": 2025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.420203300539441e-06,
+      "loss": 3.6303,
+      "step": 2026
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.3658232728621825e-06,
+      "loss": 4.207,
+      "step": 2027
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.31163562430287e-06,
+      "loss": 4.4676,
+      "step": 2028
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.257640467396365e-06,
+      "loss": 4.521,
+      "step": 2029
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.203837914277689e-06,
+      "loss": 3.9273,
+      "step": 2030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.1502280766819525e-06,
+      "loss": 4.6999,
+      "step": 2031
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.096811065944031e-06,
+      "loss": 5.132,
+      "step": 2032
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.043586992998319e-06,
+      "loss": 4.8168,
+      "step": 2033
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.990555968378531e-06,
+      "loss": 4.0264,
+      "step": 2034
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.937718102217461e-06,
+      "loss": 4.1284,
+      "step": 2035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.885073504246797e-06,
+      "loss": 3.0325,
+      "step": 2036
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.832622283796841e-06,
+      "loss": 4.2767,
+      "step": 2037
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.780364549796237e-06,
+      "loss": 5.7471,
+      "step": 2038
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.728300410771871e-06,
+      "loss": 3.8311,
+      "step": 2039
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.676429974848563e-06,
+      "loss": 5.2012,
+      "step": 2040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.624753349748869e-06,
+      "loss": 5.5337,
+      "step": 2041
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.573270642792783e-06,
+      "loss": 4.3075,
+      "step": 2042
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.521981960897638e-06,
+      "loss": 5.031,
+      "step": 2043
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.470887410577853e-06,
+      "loss": 2.6419,
+      "step": 2044
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.419987097944579e-06,
+      "loss": 5.4642,
+      "step": 2045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.369281128705673e-06,
+      "loss": 3.7059,
+      "step": 2046
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.318769608165332e-06,
+      "loss": 4.058,
+      "step": 2047
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.2684526412239765e-06,
+      "loss": 7.0315,
+      "step": 2048
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.218330332377953e-06,
+      "loss": 4.3274,
+      "step": 2049
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.1684027857193516e-06,
+      "loss": 4.9061,
+      "step": 2050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.118670104935764e-06,
+      "loss": 3.7136,
+      "step": 2051
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.069132393310184e-06,
+      "loss": 4.5402,
+      "step": 2052
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.019789753720595e-06,
+      "loss": 3.9901,
+      "step": 2053
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.970642288639916e-06,
+      "loss": 6.7703,
+      "step": 2054
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.921690100135713e-06,
+      "loss": 4.0685,
+      "step": 2055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.872933289870042e-06,
+      "loss": 4.6586,
+      "step": 2056
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.824371959099173e-06,
+      "loss": 5.1559,
+      "step": 2057
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.776006208673412e-06,
+      "loss": 5.0155,
+      "step": 2058
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.727836139036901e-06,
+      "loss": 4.5251,
+      "step": 2059
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.67986185022743e-06,
+      "loss": 4.5403,
+      "step": 2060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.632083441876124e-06,
+      "loss": 4.9371,
+      "step": 2061
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.584501013207388e-06,
+      "loss": 5.5308,
+      "step": 2062
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.537114663038578e-06,
+      "loss": 4.6071,
+      "step": 2063
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.489924489779863e-06,
+      "loss": 5.4726,
+      "step": 2064
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.442930591433992e-06,
+      "loss": 5.5484,
+      "step": 2065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.3961330655960874e-06,
+      "loss": 4.4191,
+      "step": 2066
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.349532009453439e-06,
+      "loss": 3.441,
+      "step": 2067
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.303127519785411e-06,
+      "loss": 4.4136,
+      "step": 2068
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.256919692963003e-06,
+      "loss": 5.2367,
+      "step": 2069
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.210908624948896e-06,
+      "loss": 4.0911,
+      "step": 2070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.165094411297111e-06,
+      "loss": 5.6767,
+      "step": 2071
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.119477147152851e-06,
+      "loss": 4.5841,
+      "step": 2072
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.074056927252335e-06,
+      "loss": 4.8708,
+      "step": 2073
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 5.028833845922509e-06,
+      "loss": 3.4986,
+      "step": 2074
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.983807997080925e-06,
+      "loss": 5.9033,
+      "step": 2075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.938979474235594e-06,
+      "loss": 4.8523,
+      "step": 2076
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.8943483704846475e-06,
+      "loss": 3.8246,
+      "step": 2077
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.849914778516262e-06,
+      "loss": 3.8741,
+      "step": 2078
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.805678790608414e-06,
+      "loss": 5.9754,
+      "step": 2079
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.76164049862875e-06,
+      "loss": 5.096,
+      "step": 2080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.7177999940342554e-06,
+      "loss": 4.5615,
+      "step": 2081
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6741573678712565e-06,
+      "loss": 6.3813,
+      "step": 2082
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.630712710775054e-06,
+      "loss": 4.976,
+      "step": 2083
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.587466112969907e-06,
+      "loss": 3.7464,
+      "step": 2084
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.544417664268652e-06,
+      "loss": 3.8875,
+      "step": 2085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.501567454072675e-06,
+      "loss": 4.508,
+      "step": 2086
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.458915571371647e-06,
+      "loss": 3.2686,
+      "step": 2087
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.416462104743402e-06,
+      "loss": 3.1826,
+      "step": 2088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.3742071423536255e-06,
+      "loss": 4.7067,
+      "step": 2089
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.332150771955845e-06,
+      "loss": 4.5242,
+      "step": 2090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.2902930808910945e-06,
+      "loss": 4.3622,
+      "step": 2091
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.248634156087872e-06,
+      "loss": 6.371,
+      "step": 2092
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.207174084061804e-06,
+      "loss": 3.9516,
+      "step": 2093
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.165912950915585e-06,
+      "loss": 5.953,
+      "step": 2094
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.124850842338779e-06,
+      "loss": 4.8647,
+      "step": 2095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.083987843607595e-06,
+      "loss": 4.8252,
+      "step": 2096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.043324039584728e-06,
+      "loss": 5.7428,
+      "step": 2097
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.002859514719237e-06,
+      "loss": 5.0673,
+      "step": 2098
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.962594353046279e-06,
+      "loss": 5.1344,
+      "step": 2099
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.922528638187039e-06,
+      "loss": 4.9177,
+      "step": 2100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.882662453348429e-06,
+      "loss": 4.4256,
+      "step": 2101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.8429958813230195e-06,
+      "loss": 4.8768,
+      "step": 2102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.8035290044888484e-06,
+      "loss": 5.6592,
+      "step": 2103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.764261904809219e-06,
+      "loss": 4.7214,
+      "step": 2104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.725194663832521e-06,
+      "loss": 4.2084,
+      "step": 2105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.6863273626921213e-06,
+      "loss": 3.3916,
+      "step": 2106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.6476600821061303e-06,
+      "loss": 3.0083,
+      "step": 2107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.609192902377323e-06,
+      "loss": 3.3188,
+      "step": 2108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.5709259033928188e-06,
+      "loss": 5.0073,
+      "step": 2109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.5328591646240695e-06,
+      "loss": 4.9522,
+      "step": 2110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.494992765126637e-06,
+      "loss": 5.8443,
+      "step": 2111
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.457326783540027e-06,
+      "loss": 4.5394,
+      "step": 2112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.4198612980874677e-06,
+      "loss": 5.7578,
+      "step": 2113
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.382596386575876e-06,
+      "loss": 5.3203,
+      "step": 2114
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3455321263955786e-06,
+      "loss": 3.8639,
+      "step": 2115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3086685945202366e-06,
+      "loss": 4.2976,
+      "step": 2116
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2720058675066112e-06,
+      "loss": 4.474,
+      "step": 2117
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.235544021494452e-06,
+      "loss": 4.1836,
+      "step": 2118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.199283132206332e-06,
+      "loss": 4.6268,
+      "step": 2119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.163223274947491e-06,
+      "loss": 5.0713,
+      "step": 2120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.127364524605636e-06,
+      "loss": 5.3174,
+      "step": 2121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.091706955650864e-06,
+      "loss": 5.6775,
+      "step": 2122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0562506421354496e-06,
+      "loss": 3.7339,
+      "step": 2123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0209956576937147e-06,
+      "loss": 4.8431,
+      "step": 2124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9859420755418477e-06,
+      "loss": 3.2833,
+      "step": 2125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.951089968477805e-06,
+      "loss": 5.4029,
+      "step": 2126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9164394088811107e-06,
+      "loss": 4.2694,
+      "step": 2127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.881990468712725e-06,
+      "loss": 4.8595,
+      "step": 2128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.847743219514876e-06,
+      "loss": 4.16,
+      "step": 2129
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.813697732410947e-06,
+      "loss": 2.9203,
+      "step": 2130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.779854078105304e-06,
+      "loss": 3.3691,
+      "step": 2131
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7462123268831573e-06,
+      "loss": 4.2038,
+      "step": 2132
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.7127725486104093e-06,
+      "loss": 5.523,
+      "step": 2133
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6795348127335085e-06,
+      "loss": 5.1775,
+      "step": 2134
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6464991882793278e-06,
+      "loss": 4.4898,
+      "step": 2135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6136657438549984e-06,
+      "loss": 5.3154,
+      "step": 2136
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5810345476477206e-06,
+      "loss": 3.6086,
+      "step": 2137
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5486056674247527e-06,
+      "loss": 5.3484,
+      "step": 2138
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.516379170533134e-06,
+      "loss": 5.075,
+      "step": 2139
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.48435512389964e-06,
+      "loss": 3.7034,
+      "step": 2140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4525335940305706e-06,
+      "loss": 5.8051,
+      "step": 2141
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.4209146470116627e-06,
+      "loss": 4.9804,
+      "step": 2142
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.389498348507968e-06,
+      "loss": 5.9994,
+      "step": 2143
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3582847637636072e-06,
+      "loss": 4.734,
+      "step": 2144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3272739576017945e-06,
+      "loss": 4.7118,
+      "step": 2145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.296465994424568e-06,
+      "loss": 4.3164,
+      "step": 2146
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2658609382127384e-06,
+      "loss": 4.5929,
+      "step": 2147
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.235458852525707e-06,
+      "loss": 3.64,
+      "step": 2148
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.20525980050138e-06,
+      "loss": 5.5475,
+      "step": 2149
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.175263844855979e-06,
+      "loss": 2.4801,
+      "step": 2150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1454710478839735e-06,
+      "loss": 3.4651,
+      "step": 2151
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.115881471457892e-06,
+      "loss": 4.0298,
+      "step": 2152
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.086495177028236e-06,
+      "loss": 5.2332,
+      "step": 2153
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0573122256233425e-06,
+      "loss": 4.5356,
+      "step": 2154
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.028332677849254e-06,
+      "loss": 4.5727,
+      "step": 2155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9995565938895844e-06,
+      "loss": 5.9234,
+      "step": 2156
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9709840335053965e-06,
+      "loss": 4.3102,
+      "step": 2157
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.942615056035113e-06,
+      "loss": 4.4123,
+      "step": 2158
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.914449720394329e-06,
+      "loss": 2.9795,
+      "step": 2159
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8864880850757327e-06,
+      "loss": 4.3748,
+      "step": 2160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8587302081489732e-06,
+      "loss": 2.5884,
+      "step": 2161
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8311761472605716e-06,
+      "loss": 4.0475,
+      "step": 2162
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8038259596337314e-06,
+      "loss": 4.7889,
+      "step": 2163
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7766797020682845e-06,
+      "loss": 3.8434,
+      "step": 2164
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7497374309405346e-06,
+      "loss": 5.4274,
+      "step": 2165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7229992022031682e-06,
+      "loss": 2.9643,
+      "step": 2166
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6964650713851226e-06,
+      "loss": 4.6313,
+      "step": 2167
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6701350935914518e-06,
+      "loss": 4.4161,
+      "step": 2168
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6440093235032484e-06,
+      "loss": 4.7646,
+      "step": 2169
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6180878153775113e-06,
+      "loss": 5.1919,
+      "step": 2170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.592370623047046e-06,
+      "loss": 4.3531,
+      "step": 2171
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5668577999203404e-06,
+      "loss": 5.458,
+      "step": 2172
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5415493989814457e-06,
+      "loss": 4.627,
+      "step": 2173
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5164454727898737e-06,
+      "loss": 4.2468,
+      "step": 2174
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4915460734805097e-06,
+      "loss": 4.1341,
+      "step": 2175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4668512527634681e-06,
+      "loss": 5.041,
+      "step": 2176
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4423610619240025e-06,
+      "loss": 5.9214,
+      "step": 2177
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4180755518224175e-06,
+      "loss": 5.1788,
+      "step": 2178
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3939947728939251e-06,
+      "loss": 4.3453,
+      "step": 2179
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3701187751485656e-06,
+      "loss": 4.9155,
+      "step": 2180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3464476081711085e-06,
+      "loss": 5.3533,
+      "step": 2181
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.3229813211209197e-06,
+      "loss": 4.4547,
+      "step": 2182
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2997199627319045e-06,
+      "loss": 4.0111,
+      "step": 2183
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2766635813123317e-06,
+      "loss": 3.661,
+      "step": 2184
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2538122247448325e-06,
+      "loss": 6.2592,
+      "step": 2185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.231165940486234e-06,
+      "loss": 4.5362,
+      "step": 2186
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2087247755674601e-06,
+      "loss": 4.8537,
+      "step": 2187
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.186488776593475e-06,
+      "loss": 5.6022,
+      "step": 2188
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1644579897431507e-06,
+      "loss": 2.7073,
+      "step": 2189
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.1426324607691774e-06,
+      "loss": 5.184,
+      "step": 2190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.121012234997998e-06,
+      "loss": 5.143,
+      "step": 2191
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0995973573296403e-06,
+      "loss": 2.8722,
+      "step": 2192
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0783878722377295e-06,
+      "loss": 4.9533,
+      "step": 2193
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0573838237692979e-06,
+      "loss": 4.3358,
+      "step": 2194
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.036585255544764e-06,
+      "loss": 5.5152,
+      "step": 2195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0159922107577879e-06,
+      "loss": 5.4482,
+      "step": 2196
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.95604732175215e-07,
+      "loss": 3.7224,
+      "step": 2197
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.754228621369988e-07,
+      "loss": 5.1601,
+      "step": 2198
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.554466425560792e-07,
+      "loss": 5.3431,
+      "step": 2199
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.356761149182713e-07,
+      "loss": 3.2655,
+      "step": 2200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.161113202822868e-07,
+      "loss": 4.4248,
+      "step": 2201
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.967522992795241e-07,
+      "loss": 5.503,
+      "step": 2202
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.775990921140564e-07,
+      "loss": 3.6614,
+      "step": 2203
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.586517385625437e-07,
+      "loss": 6.0496,
+      "step": 2204
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.399102779741319e-07,
+      "loss": 5.4007,
+      "step": 2205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.213747492703427e-07,
+      "loss": 4.5903,
+      "step": 2206
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.030451909450842e-07,
+      "loss": 6.5981,
+      "step": 2207
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.849216410644622e-07,
+      "loss": 5.2626,
+      "step": 2208
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.670041372667802e-07,
+      "loss": 5.2819,
+      "step": 2209
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.492927167624287e-07,
+      "loss": 5.5867,
+      "step": 2210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.317874163338068e-07,
+      "loss": 1.912,
+      "step": 2211
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.144882723352342e-07,
+      "loss": 5.339,
+      "step": 2212
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.973953206929506e-07,
+      "loss": 4.8062,
+      "step": 2213
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.805085969049164e-07,
+      "loss": 3.9781,
+      "step": 2214
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.638281360408339e-07,
+      "loss": 5.068,
+      "step": 2215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.473539727420597e-07,
+      "loss": 3.7118,
+      "step": 2216
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.31086141221493e-07,
+      "loss": 2.6705,
+      "step": 2217
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.150246752635535e-07,
+      "loss": 5.4056,
+      "step": 2218
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.991696082240816e-07,
+      "loss": 5.3232,
+      "step": 2219
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.835209730302715e-07,
+      "loss": 4.8927,
+      "step": 2220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.68078802180616e-07,
+      "loss": 5.2284,
+      "step": 2221
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.528431277448177e-07,
+      "loss": 4.2192,
+      "step": 2222
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.37813981363744e-07,
+      "loss": 5.2323,
+      "step": 2223
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.229913942493503e-07,
+      "loss": 6.0174,
+      "step": 2224
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.083753971846239e-07,
+      "loss": 4.9155,
+      "step": 2225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.939660205235064e-07,
+      "loss": 5.3971,
+      "step": 2226
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.797632941908492e-07,
+      "loss": 5.2051,
+      "step": 2227
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.657672476823138e-07,
+      "loss": 4.8251,
+      "step": 2228
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.519779100643606e-07,
+      "loss": 4.8043,
+      "step": 2229
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.383953099741711e-07,
+      "loss": 3.4133,
+      "step": 2230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.250194756195591e-07,
+      "loss": 4.1217,
+      "step": 2231
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.118504347789487e-07,
+      "loss": 4.4134,
+      "step": 2232
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.988882148012962e-07,
+      "loss": 2.8254,
+      "step": 2233
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.8613284260605733e-07,
+      "loss": 3.9806,
+      "step": 2234
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.7358434468308667e-07,
+      "loss": 3.0739,
+      "step": 2235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.612427470926383e-07,
+      "loss": 3.9087,
+      "step": 2236
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.4910807546525427e-07,
+      "loss": 5.5413,
+      "step": 2237
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.3718035500177604e-07,
+      "loss": 5.8267,
+      "step": 2238
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2545961047322213e-07,
+      "loss": 5.1375,
+      "step": 2239
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.139458662207995e-07,
+      "loss": 4.2522,
+      "step": 2240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.02639146155792e-07,
+      "loss": 4.8625,
+      "step": 2241
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9153947375958333e-07,
+      "loss": 4.2742,
+      "step": 2242
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.8064687208354534e-07,
+      "loss": 4.2421,
+      "step": 2243
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6996136374901613e-07,
+      "loss": 4.1435,
+      "step": 2244
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.594829709472446e-07,
+      "loss": 4.141,
+      "step": 2245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.492117154393681e-07,
+      "loss": 4.3602,
+      "step": 2246
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.391476185563235e-07,
+      "loss": 3.2943,
+      "step": 2247
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.2929070119884767e-07,
+      "loss": 4.4394,
+      "step": 2248
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1964098383742137e-07,
+      "loss": 3.4649,
+      "step": 2249
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.1019848651219199e-07,
+      "loss": 4.2044,
+      "step": 2250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0096322883298435e-07,
+      "loss": 4.8123,
+      "step": 2251
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919352299792343e-07,
+      "loss": 3.5186,
+      "step": 2252
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.8311450869993307e-07,
+      "loss": 5.1012,
+      "step": 2253
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7450108331362736e-07,
+      "loss": 3.726,
+      "step": 2254
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6609497170834155e-07,
+      "loss": 4.5189,
+      "step": 2255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5789619134157774e-07,
+      "loss": 4.6145,
+      "step": 2256
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4990475924023807e-07,
+      "loss": 6.1374,
+      "step": 2257
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4212069200063573e-07,
+      "loss": 3.4211,
+      "step": 2258
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.345440057884062e-07,
+      "loss": 3.805,
+      "step": 2259
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.271747163385295e-07,
+      "loss": 5.0087,
+      "step": 2260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.200128389552857e-07,
+      "loss": 3.4076,
+      "step": 2261
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.13058388512155e-07,
+      "loss": 4.2768,
+      "step": 2262
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0631137945189551e-07,
+      "loss": 4.6649,
+      "step": 2263
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.977182578643219e-08,
+      "loss": 5.2891,
+      "step": 2264
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.343974109685682e-08,
+      "loss": 4.7628,
+      "step": 2265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.731513853340589e-08,
+      "loss": 5.2521,
+      "step": 2266
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.139803081540498e-08,
+      "loss": 5.4228,
+      "step": 2267
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.568843023130213e-08,
+      "loss": 5.9154,
+      "step": 2268
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.018634863856788e-08,
+      "loss": 5.089,
+      "step": 2269
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.489179746370644e-08,
+      "loss": 3.3637,
+      "step": 2270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.980478770224452e-08,
+      "loss": 4.1639,
+      "step": 2271
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.492532991868693e-08,
+      "loss": 4.7366,
+      "step": 2272
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.025343424651663e-08,
+      "loss": 4.1451,
+      "step": 2273
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.578911038813916e-08,
+      "loss": 4.6922,
+      "step": 2274
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.153236761488266e-08,
+      "loss": 4.532,
+      "step": 2275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.7483214767009e-08,
+      "loss": 4.5325,
+      "step": 2276
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.364166025362492e-08,
+      "loss": 4.5565,
+      "step": 2277
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0007712052737556e-08,
+      "loss": 4.9796,
+      "step": 2278
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.6581377711176747e-08,
+      "loss": 5.1681,
+      "step": 2279
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.3362664344639407e-08,
+      "loss": 4.3366,
+      "step": 2280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.0351578637611833e-08,
+      "loss": 4.7785,
+      "step": 2281
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.7548126843425215e-08,
+      "loss": 3.8582,
+      "step": 2282
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4952314784144607e-08,
+      "loss": 5.5997,
+      "step": 2283
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.2564147850679942e-08,
+      "loss": 4.8249,
+      "step": 2284
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.0383631002686134e-08,
+      "loss": 4.5705,
+      "step": 2285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 8.41076876856306e-09,
+      "loss": 3.6896,
+      "step": 2286
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 6.64556524547777e-09,
+      "loss": 6.2818,
+      "step": 2287
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.088024099353383e-09,
+      "loss": 4.5365,
+      "step": 2288
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.738148564813582e-09,
+      "loss": 4.7985,
+      "step": 2289
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.5959414452270124e-09,
+      "loss": 5.4271,
+      "step": 2290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.6614051127072928e-09,
+      "loss": 4.7306,
+      "step": 2291
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.345415080463936e-10,
+      "loss": 4.6008,
+      "step": 2292
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.1535214075905106e-10,
+      "loss": 5.1314,
+      "step": 2293
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.038380891049684e-10,
+      "loss": 3.8032,
+      "step": 2294
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0,
+      "loss": 4.835,
+      "step": 2295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.038380891049684e-10,
+      "loss": 4.8936,
+      "step": 2296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.1535214075905106e-10,
+      "loss": 4.4959,
+      "step": 2297
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.345415080463936e-10,
+      "loss": 5.0679,
+      "step": 2298
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6614051127072928e-09,
+      "loss": 4.3543,
+      "step": 2299
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5959414452270124e-09,
+      "loss": 5.7222,
+      "step": 2300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.738148564813582e-09,
+      "loss": 5.7231,
+      "step": 2301
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.088024099353383e-09,
+      "loss": 4.5728,
+      "step": 2302
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.64556524547777e-09,
+      "loss": 4.3249,
+      "step": 2303
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.41076876856306e-09,
+      "loss": 6.0688,
+      "step": 2304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0383631002686134e-08,
+      "loss": 3.9971,
+      "step": 2305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2564147850679942e-08,
+      "loss": 3.1084,
+      "step": 2306
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4952314784144607e-08,
+      "loss": 3.7402,
+      "step": 2307
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7548126843425215e-08,
+      "loss": 5.4944,
+      "step": 2308
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0351578637611833e-08,
+      "loss": 5.3342,
+      "step": 2309
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3362664344639407e-08,
+      "loss": 4.8981,
+      "step": 2310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6581377711176747e-08,
+      "loss": 3.312,
+      "step": 2311
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0007712052737556e-08,
+      "loss": 5.3513,
+      "step": 2312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.364166025362492e-08,
+      "loss": 4.6219,
+      "step": 2313
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.7483214767009e-08,
+      "loss": 2.9824,
+      "step": 2314
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.153236761488266e-08,
+      "loss": 5.5241,
+      "step": 2315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.5789110388128055e-08,
+      "loss": 5.1638,
+      "step": 2316
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.025343424651663e-08,
+      "loss": 4.993,
+      "step": 2317
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.492532991868693e-08,
+      "loss": 5.3677,
+      "step": 2318
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.980478770224452e-08,
+      "loss": 5.2659,
+      "step": 2319
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.489179746370644e-08,
+      "loss": 4.3444,
+      "step": 2320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.018634863856788e-08,
+      "loss": 5.8409,
+      "step": 2321
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.568843023130213e-08,
+      "loss": 3.8386,
+      "step": 2322
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.139803081540498e-08,
+      "loss": 4.4621,
+      "step": 2323
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.731513853340589e-08,
+      "loss": 3.6424,
+      "step": 2324
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.343974109685682e-08,
+      "loss": 5.034,
+      "step": 2325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.977182578643219e-08,
+      "loss": 6.0143,
+      "step": 2326
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0631137945189551e-07,
+      "loss": 4.9618,
+      "step": 2327
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.13058388512155e-07,
+      "loss": 4.1779,
+      "step": 2328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.200128389552746e-07,
+      "loss": 4.0761,
+      "step": 2329
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.271747163385295e-07,
+      "loss": 5.1628,
+      "step": 2330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.345440057884062e-07,
+      "loss": 5.4052,
+      "step": 2331
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.421206920006246e-07,
+      "loss": 3.9752,
+      "step": 2332
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4990475924023807e-07,
+      "loss": 4.6869,
+      "step": 2333
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5789619134156662e-07,
+      "loss": 4.6865,
+      "step": 2334
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6609497170834155e-07,
+      "loss": 4.0655,
+      "step": 2335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7450108331362736e-07,
+      "loss": 4.6544,
+      "step": 2336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8311450869993307e-07,
+      "loss": 4.1208,
+      "step": 2337
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.919352299792343e-07,
+      "loss": 5.3791,
+      "step": 2338
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0096322883298435e-07,
+      "loss": 4.7024,
+      "step": 2339
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1019848651219199e-07,
+      "loss": 3.5452,
+      "step": 2340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1964098383742137e-07,
+      "loss": 4.3767,
+      "step": 2341
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.2929070119884767e-07,
+      "loss": 5.5294,
+      "step": 2342
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.391476185563235e-07,
+      "loss": 5.6141,
+      "step": 2343
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.492117154393681e-07,
+      "loss": 3.1481,
+      "step": 2344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.594829709472446e-07,
+      "loss": 3.6464,
+      "step": 2345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6996136374901613e-07,
+      "loss": 6.6393,
+      "step": 2346
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8064687208354534e-07,
+      "loss": 5.0408,
+      "step": 2347
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9153947375958333e-07,
+      "loss": 4.5508,
+      "step": 2348
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.02639146155792e-07,
+      "loss": 4.4195,
+      "step": 2349
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1394586622078836e-07,
+      "loss": 4.77,
+      "step": 2350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2545961047322213e-07,
+      "loss": 5.051,
+      "step": 2351
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3718035500177604e-07,
+      "loss": 5.1866,
+      "step": 2352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.4910807546525427e-07,
+      "loss": 4.606,
+      "step": 2353
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.612427470926383e-07,
+      "loss": 4.8627,
+      "step": 2354
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.7358434468308667e-07,
+      "loss": 4.6127,
+      "step": 2355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.8613284260605733e-07,
+      "loss": 3.8019,
+      "step": 2356
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.988882148012962e-07,
+      "loss": 5.2997,
+      "step": 2357
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.118504347789487e-07,
+      "loss": 5.5139,
+      "step": 2358
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.250194756195591e-07,
+      "loss": 4.9092,
+      "step": 2359
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.383953099741711e-07,
+      "loss": 5.43,
+      "step": 2360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.519779100643606e-07,
+      "loss": 4.8188,
+      "step": 2361
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.657672476823027e-07,
+      "loss": 5.6503,
+      "step": 2362
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.797632941908381e-07,
+      "loss": 3.8202,
+      "step": 2363
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.939660205235064e-07,
+      "loss": 5.7599,
+      "step": 2364
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.083753971846239e-07,
+      "loss": 3.3385,
+      "step": 2365
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.229913942493503e-07,
+      "loss": 4.6566,
+      "step": 2366
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.37813981363744e-07,
+      "loss": 5.2873,
+      "step": 2367
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.528431277448177e-07,
+      "loss": 6.488,
+      "step": 2368
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.68078802180616e-07,
+      "loss": 4.0493,
+      "step": 2369
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.835209730302827e-07,
+      "loss": 5.4589,
+      "step": 2370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.991696082240927e-07,
+      "loss": 3.8813,
+      "step": 2371
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.150246752635535e-07,
+      "loss": 5.3425,
+      "step": 2372
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.31086141221493e-07,
+      "loss": 4.6848,
+      "step": 2373
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.473539727420486e-07,
+      "loss": 4.8181,
+      "step": 2374
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.638281360408339e-07,
+      "loss": 4.9947,
+      "step": 2375
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.805085969049052e-07,
+      "loss": 6.2826,
+      "step": 2376
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 6.973953206929396e-07,
+      "loss": 4.1566,
+      "step": 2377
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.144882723352342e-07,
+      "loss": 3.6647,
+      "step": 2378
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.317874163337957e-07,
+      "loss": 5.0677,
+      "step": 2379
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.492927167624287e-07,
+      "loss": 4.9958,
+      "step": 2380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.670041372667802e-07,
+      "loss": 2.8653,
+      "step": 2381
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.849216410644622e-07,
+      "loss": 5.4647,
+      "step": 2382
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.030451909450731e-07,
+      "loss": 5.5368,
+      "step": 2383
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.213747492703427e-07,
+      "loss": 5.845,
+      "step": 2384
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.399102779741319e-07,
+      "loss": 3.9383,
+      "step": 2385
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.586517385625548e-07,
+      "loss": 4.7755,
+      "step": 2386
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.775990921140564e-07,
+      "loss": 5.4912,
+      "step": 2387
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 8.967522992795129e-07,
+      "loss": 4.076,
+      "step": 2388
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.161113202822757e-07,
+      "loss": 6.1826,
+      "step": 2389
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.356761149182713e-07,
+      "loss": 4.2974,
+      "step": 2390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.554466425560682e-07,
+      "loss": 5.5253,
+      "step": 2391
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.754228621369988e-07,
+      "loss": 5.1111,
+      "step": 2392
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.95604732175215e-07,
+      "loss": 4.7066,
+      "step": 2393
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0159922107577769e-06,
+      "loss": 3.9218,
+      "step": 2394
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.036585255544753e-06,
+      "loss": 4.915,
+      "step": 2395
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0573838237692979e-06,
+      "loss": 5.0847,
+      "step": 2396
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0783878722377183e-06,
+      "loss": 4.3831,
+      "step": 2397
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.0995973573296293e-06,
+      "loss": 4.5108,
+      "step": 2398
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1210122349979868e-06,
+      "loss": 5.4447,
+      "step": 2399
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1426324607691774e-06,
+      "loss": 5.0833,
+      "step": 2400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.1644579897431507e-06,
+      "loss": 5.0514,
+      "step": 2401
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.186488776593475e-06,
+      "loss": 3.8297,
+      "step": 2402
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2087247755674601e-06,
+      "loss": 4.075,
+      "step": 2403
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.231165940486223e-06,
+      "loss": 3.5666,
+      "step": 2404
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2538122247448325e-06,
+      "loss": 4.4586,
+      "step": 2405
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2766635813123317e-06,
+      "loss": 5.2005,
+      "step": 2406
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.2997199627318934e-06,
+      "loss": 4.9904,
+      "step": 2407
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3229813211209197e-06,
+      "loss": 4.9902,
+      "step": 2408
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3464476081711085e-06,
+      "loss": 4.633,
+      "step": 2409
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3701187751485656e-06,
+      "loss": 3.6083,
+      "step": 2410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.3939947728939251e-06,
+      "loss": 5.1942,
+      "step": 2411
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4180755518224065e-06,
+      "loss": 5.5261,
+      "step": 2412
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4423610619239914e-06,
+      "loss": 5.4346,
+      "step": 2413
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4668512527634571e-06,
+      "loss": 5.1052,
+      "step": 2414
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.4915460734804987e-06,
+      "loss": 5.4106,
+      "step": 2415
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5164454727898737e-06,
+      "loss": 4.2392,
+      "step": 2416
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5415493989814457e-06,
+      "loss": 5.52,
+      "step": 2417
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.5668577999203516e-06,
+      "loss": 4.4651,
+      "step": 2418
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.592370623047046e-06,
+      "loss": 4.3293,
+      "step": 2419
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6180878153775113e-06,
+      "loss": 4.8642,
+      "step": 2420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6440093235032372e-06,
+      "loss": 4.7183,
+      "step": 2421
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6701350935914406e-06,
+      "loss": 7.232,
+      "step": 2422
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.6964650713851226e-06,
+      "loss": 4.4802,
+      "step": 2423
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7229992022031682e-06,
+      "loss": 3.4936,
+      "step": 2424
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7497374309405346e-06,
+      "loss": 2.8331,
+      "step": 2425
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.7766797020682845e-06,
+      "loss": 4.9828,
+      "step": 2426
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8038259596337314e-06,
+      "loss": 3.3585,
+      "step": 2427
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8311761472605604e-06,
+      "loss": 4.0684,
+      "step": 2428
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8587302081489622e-06,
+      "loss": 3.9005,
+      "step": 2429
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.8864880850757105e-06,
+      "loss": 5.285,
+      "step": 2430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9144497203943177e-06,
+      "loss": 4.4009,
+      "step": 2431
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942615056035113e-06,
+      "loss": 5.1744,
+      "step": 2432
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9709840335054075e-06,
+      "loss": 4.2107,
+      "step": 2433
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9995565938895844e-06,
+      "loss": 4.3109,
+      "step": 2434
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.028332677849254e-06,
+      "loss": 4.7619,
+      "step": 2435
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.0573122256233425e-06,
+      "loss": 2.7344,
+      "step": 2436
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.086495177028236e-06,
+      "loss": 5.4782,
+      "step": 2437
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.115881471457881e-06,
+      "loss": 3.8399,
+      "step": 2438
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1454710478839735e-06,
+      "loss": 3.9038,
+      "step": 2439
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.175263844855968e-06,
+      "loss": 5.0394,
+      "step": 2440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.205259800501369e-06,
+      "loss": 3.98,
+      "step": 2441
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.235458852525707e-06,
+      "loss": 3.6053,
+      "step": 2442
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.265860938212727e-06,
+      "loss": 5.1482,
+      "step": 2443
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.296465994424546e-06,
+      "loss": 5.3982,
+      "step": 2444
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.327273957601772e-06,
+      "loss": 5.1484,
+      "step": 2445
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.3582847637635962e-06,
+      "loss": 6.649,
+      "step": 2446
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.389498348507946e-06,
+      "loss": 4.8341,
+      "step": 2447
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.420914647011674e-06,
+      "loss": 3.8165,
+      "step": 2448
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.4525335940305706e-06,
+      "loss": 4.8343,
+      "step": 2449
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.48435512389964e-06,
+      "loss": 5.5229,
+      "step": 2450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.516379170533134e-06,
+      "loss": 4.8422,
+      "step": 2451
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5486056674247417e-06,
+      "loss": 4.6471,
+      "step": 2452
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.5810345476477206e-06,
+      "loss": 4.7779,
+      "step": 2453
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.613665743854987e-06,
+      "loss": 6.0536,
+      "step": 2454
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6464991882793278e-06,
+      "loss": 4.443,
+      "step": 2455
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.6795348127335085e-06,
+      "loss": 4.9689,
+      "step": 2456
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7127725486103982e-06,
+      "loss": 4.8262,
+      "step": 2457
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.7462123268831463e-06,
+      "loss": 5.1052,
+      "step": 2458
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.779854078105293e-06,
+      "loss": 2.724,
+      "step": 2459
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.813697732410925e-06,
+      "loss": 3.7747,
+      "step": 2460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.8477432195148533e-06,
+      "loss": 4.3055,
+      "step": 2461
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.881990468712714e-06,
+      "loss": 5.0974,
+      "step": 2462
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9164394088811107e-06,
+      "loss": 3.8852,
+      "step": 2463
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.951089968477816e-06,
+      "loss": 4.4982,
+      "step": 2464
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9859420755418588e-06,
+      "loss": 6.0171,
+      "step": 2465
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0209956576937147e-06,
+      "loss": 6.0219,
+      "step": 2466
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.0562506421354386e-06,
+      "loss": 4.0239,
+      "step": 2467
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.091706955650853e-06,
+      "loss": 4.46,
+      "step": 2468
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.1273645246056247e-06,
+      "loss": 3.4766,
+      "step": 2469
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.16322327494748e-06,
+      "loss": 5.0189,
+      "step": 2470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.199283132206332e-06,
+      "loss": 3.8458,
+      "step": 2471
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.235544021494452e-06,
+      "loss": 6.8556,
+      "step": 2472
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.2720058675066002e-06,
+      "loss": 5.4413,
+      "step": 2473
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3086685945202366e-06,
+      "loss": 4.566,
+      "step": 2474
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3455321263955786e-06,
+      "loss": 4.5476,
+      "step": 2475
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3825963865758535e-06,
+      "loss": 6.7054,
+      "step": 2476
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.4198612980874456e-06,
+      "loss": 3.4748,
+      "step": 2477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.457326783540005e-06,
+      "loss": 5.5759,
+      "step": 2478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.494992765126637e-06,
+      "loss": 5.2826,
+      "step": 2479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.5328591646240805e-06,
+      "loss": 5.6765,
+      "step": 2480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.5709259033928188e-06,
+      "loss": 3.1822,
+      "step": 2481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.609192902377323e-06,
+      "loss": 5.061,
+      "step": 2482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.6476600821061413e-06,
+      "loss": 5.5412,
+      "step": 2483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.6863273626921213e-06,
+      "loss": 4.5117,
+      "step": 2484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.725194663832521e-06,
+      "loss": 6.7872,
+      "step": 2485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.764261904809219e-06,
+      "loss": 4.6472,
+      "step": 2486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.8035290044888484e-06,
+      "loss": 4.5692,
+      "step": 2487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.8429958813230195e-06,
+      "loss": 4.8181,
+      "step": 2488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.882662453348418e-06,
+      "loss": 2.5259,
+      "step": 2489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.922528638187028e-06,
+      "loss": 5.339,
+      "step": 2490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.962594353046268e-06,
+      "loss": 5.0804,
+      "step": 2491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.0028595147192145e-06,
+      "loss": 4.4576,
+      "step": 2492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.043324039584717e-06,
+      "loss": 6.0624,
+      "step": 2493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.083987843607573e-06,
+      "loss": 5.0136,
+      "step": 2494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.124850842338768e-06,
+      "loss": 4.2571,
+      "step": 2495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.165912950915596e-06,
+      "loss": 4.2634,
+      "step": 2496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.207174084061804e-06,
+      "loss": 3.7592,
+      "step": 2497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.248634156087872e-06,
+      "loss": 4.7246,
+      "step": 2498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.2902930808911055e-06,
+      "loss": 4.5408,
+      "step": 2499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.332150771955834e-06,
+      "loss": 5.8806,
+      "step": 2500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3742071423536255e-06,
+      "loss": 5.187,
+      "step": 2501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.4164621047433906e-06,
+      "loss": 3.0868,
+      "step": 2502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.458915571371647e-06,
+      "loss": 4.6339,
+      "step": 2503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.501567454072664e-06,
+      "loss": 5.6807,
+      "step": 2504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5444176642686405e-06,
+      "loss": 3.9792,
+      "step": 2505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5874661129698955e-06,
+      "loss": 3.2958,
+      "step": 2506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.630712710775043e-06,
+      "loss": 5.4688,
+      "step": 2507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.6741573678712345e-06,
+      "loss": 3.1102,
+      "step": 2508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.717799994034233e-06,
+      "loss": 4.4152,
+      "step": 2509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.761640498628728e-06,
+      "loss": 4.863,
+      "step": 2510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.805678790608414e-06,
+      "loss": 3.993,
+      "step": 2511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.849914778516262e-06,
+      "loss": 4.8934,
+      "step": 2512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8943483704846475e-06,
+      "loss": 5.9344,
+      "step": 2513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.938979474235583e-06,
+      "loss": 3.597,
+      "step": 2514
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.983807997080925e-06,
+      "loss": 5.1881,
+      "step": 2515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.028833845922498e-06,
+      "loss": 3.8855,
+      "step": 2516
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.074056927252335e-06,
+      "loss": 3.6498,
+      "step": 2517
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.119477147152851e-06,
+      "loss": 4.0739,
+      "step": 2518
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.1650944112971e-06,
+      "loss": 4.5994,
+      "step": 2519
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.210908624948885e-06,
+      "loss": 4.684,
+      "step": 2520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.256919692962991e-06,
+      "loss": 4.9968,
+      "step": 2521
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.3031275197854e-06,
+      "loss": 4.4724,
+      "step": 2522
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.349532009453439e-06,
+      "loss": 4.2923,
+      "step": 2523
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.396133065596065e-06,
+      "loss": 4.1987,
+      "step": 2524
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.44293059143397e-06,
+      "loss": 3.3054,
+      "step": 2525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.489924489779852e-06,
+      "loss": 4.2708,
+      "step": 2526
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.537114663038556e-06,
+      "loss": 4.4541,
+      "step": 2527
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.584501013207399e-06,
+      "loss": 4.9219,
+      "step": 2528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.632083441876135e-06,
+      "loss": 4.2062,
+      "step": 2529
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.679861850227419e-06,
+      "loss": 7.0292,
+      "step": 2530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.727836139036901e-06,
+      "loss": 3.4727,
+      "step": 2531
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.776006208673412e-06,
+      "loss": 4.0892,
+      "step": 2532
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.824371959099173e-06,
+      "loss": 4.9125,
+      "step": 2533
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.872933289870031e-06,
+      "loss": 5.7099,
+      "step": 2534
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.921690100135713e-06,
+      "loss": 4.9222,
+      "step": 2535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.970642288639905e-06,
+      "loss": 3.9383,
+      "step": 2536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.0197897537205836e-06,
+      "loss": 5.2261,
+      "step": 2537
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.069132393310184e-06,
+      "loss": 3.8727,
+      "step": 2538
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.118670104935753e-06,
+      "loss": 5.1098,
+      "step": 2539
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.1684027857193295e-06,
+      "loss": 3.841,
+      "step": 2540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.218330332377931e-06,
+      "loss": 4.5594,
+      "step": 2541
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.2684526412239655e-06,
+      "loss": 5.9868,
+      "step": 2542
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.318769608165309e-06,
+      "loss": 4.825,
+      "step": 2543
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.369281128705673e-06,
+      "loss": 4.4538,
+      "step": 2544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.419987097944591e-06,
+      "loss": 5.2567,
+      "step": 2545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.470887410577842e-06,
+      "loss": 4.3997,
+      "step": 2546
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.521981960897649e-06,
+      "loss": 4.3897,
+      "step": 2547
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.573270642792772e-06,
+      "loss": 4.0652,
+      "step": 2548
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.624753349748858e-06,
+      "loss": 2.7186,
+      "step": 2549
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.676429974848552e-06,
+      "loss": 3.8147,
+      "step": 2550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.728300410771871e-06,
+      "loss": 4.2454,
+      "step": 2551
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.780364549796226e-06,
+      "loss": 3.8055,
+      "step": 2552
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.83262228379683e-06,
+      "loss": 3.1384,
+      "step": 2553
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.885073504246786e-06,
+      "loss": 3.9678,
+      "step": 2554
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.93771810221745e-06,
+      "loss": 3.5046,
+      "step": 2555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 6.990555968378509e-06,
+      "loss": 4.5509,
+      "step": 2556
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.043586992998297e-06,
+      "loss": 4.5367,
+      "step": 2557
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.09681106594402e-06,
+      "loss": 5.185,
+      "step": 2558
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.1502280766819305e-06,
+      "loss": 5.1273,
+      "step": 2559
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.2038379142777e-06,
+      "loss": 4.3693,
+      "step": 2560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.257640467396354e-06,
+      "loss": 4.1159,
+      "step": 2561
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.311635624302881e-06,
+      "loss": 5.1741,
+      "step": 2562
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.3658232728621825e-06,
+      "loss": 4.8209,
+      "step": 2563
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.42020330053943e-06,
+      "loss": 4.5482,
+      "step": 2564
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.474775594400252e-06,
+      "loss": 4.6464,
+      "step": 2565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.529540041110972e-06,
+      "loss": 3.1294,
+      "step": 2566
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.584496526938922e-06,
+      "loss": 4.4068,
+      "step": 2567
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.639644937752555e-06,
+      "loss": 4.9245,
+      "step": 2568
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.694985159021761e-06,
+      "loss": 4.8827,
+      "step": 2569
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.750517075818054e-06,
+      "loss": 5.3953,
+      "step": 2570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.806240572814927e-06,
+      "loss": 4.9477,
+      "step": 2571
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.862155534287908e-06,
+      "loss": 4.7756,
+      "step": 2572
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.918261844114971e-06,
+      "loss": 4.5243,
+      "step": 2573
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.974559385776647e-06,
+      "loss": 4.0419,
+      "step": 2574
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.03104804235636e-06,
+      "loss": 4.2768,
+      "step": 2575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.08772769654068e-06,
+      "loss": 4.6548,
+      "step": 2576
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.144598230619371e-06,
+      "loss": 4.4274,
+      "step": 2577
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.20165952648596e-06,
+      "loss": 5.3431,
+      "step": 2578
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.258911465637687e-06,
+      "loss": 3.9674,
+      "step": 2579
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.316353929175924e-06,
+      "loss": 5.829,
+      "step": 2580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.373986797806365e-06,
+      "loss": 4.6887,
+      "step": 2581
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.431809951839253e-06,
+      "loss": 4.6545,
+      "step": 2582
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.489823271189701e-06,
+      "loss": 4.6776,
+      "step": 2583
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.548026635377848e-06,
+      "loss": 4.164,
+      "step": 2584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.606419923529163e-06,
+      "loss": 5.7137,
+      "step": 2585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.665003014374695e-06,
+      "loss": 5.54,
+      "step": 2586
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.723775786251343e-06,
+      "loss": 4.8683,
+      "step": 2587
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.782738117102051e-06,
+      "loss": 4.0684,
+      "step": 2588
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.841889884476107e-06,
+      "loss": 3.9903,
+      "step": 2589
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.90123096552935e-06,
+      "loss": 3.6054,
+      "step": 2590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.960761237024539e-06,
+      "loss": 4.521,
+      "step": 2591
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.020480575331491e-06,
+      "loss": 4.5903,
+      "step": 2592
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.08038885642728e-06,
+      "loss": 4.3688,
+      "step": 2593
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.14048595589675e-06,
+      "loss": 4.8275,
+      "step": 2594
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.200771748932524e-06,
+      "loss": 5.5257,
+      "step": 2595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.261246110335365e-06,
+      "loss": 5.1233,
+      "step": 2596
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.32190891451441e-06,
+      "loss": 5.8129,
+      "step": 2597
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.382760035487526e-06,
+      "loss": 4.6509,
+      "step": 2598
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.443799346881388e-06,
+      "loss": 5.7049,
+      "step": 2599
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505026721931887e-06,
+      "loss": 4.4536,
+      "step": 2600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.566442033484368e-06,
+      "loss": 5.1997,
+      "step": 2601
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.628045153993825e-06,
+      "loss": 4.1916,
+      "step": 2602
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.689835955525295e-06,
+      "loss": 5.5746,
+      "step": 2603
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.751814309753981e-06,
+      "loss": 4.728,
+      "step": 2604
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.813980087965613e-06,
+      "loss": 5.1295,
+      "step": 2605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.87633316105665e-06,
+      "loss": 4.7476,
+      "step": 2606
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.938873399534654e-06,
+      "loss": 5.2932,
+      "step": 2607
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0001600673518474e-05,
+      "loss": 4.8458,
+      "step": 2608
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0064514852738393e-05,
+      "loss": 5.435,
+      "step": 2609
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.012761580653675e-05,
+      "loss": 4.0145,
+      "step": 2610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0190903403867858e-05,
+      "loss": 4.5401,
+      "step": 2611
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0254377513298453e-05,
+      "loss": 5.3455,
+      "step": 2612
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0318038003007913e-05,
+      "loss": 4.3332,
+      "step": 2613
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0381884740788616e-05,
+      "loss": 5.9193,
+      "step": 2614
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0445917594046062e-05,
+      "loss": 3.2725,
+      "step": 2615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0510136429799278e-05,
+      "loss": 4.7185,
+      "step": 2616
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0574541114681037e-05,
+      "loss": 4.6564,
+      "step": 2617
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0639131514938128e-05,
+      "loss": 6.0657,
+      "step": 2618
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0703907496431731e-05,
+      "loss": 4.5306,
+      "step": 2619
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.076886892463752e-05,
+      "loss": 4.7937,
+      "step": 2620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0834015664646102e-05,
+      "loss": 4.9201,
+      "step": 2621
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0899347581163189e-05,
+      "loss": 4.2862,
+      "step": 2622
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.0964864538509989e-05,
+      "loss": 4.868,
+      "step": 2623
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.103056640062341e-05,
+      "loss": 3.747,
+      "step": 2624
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1096453031056264e-05,
+      "loss": 3.6695,
+      "step": 2625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.116252429297776e-05,
+      "loss": 3.5696,
+      "step": 2626
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1228780049173626e-05,
+      "loss": 7.4645,
+      "step": 2627
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1295220162046417e-05,
+      "loss": 5.811,
+      "step": 2628
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1361844493615847e-05,
+      "loss": 4.5456,
+      "step": 2629
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1428652905519088e-05,
+      "loss": 4.16,
+      "step": 2630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1495645259010956e-05,
+      "loss": 5.9578,
+      "step": 2631
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1562821414964298e-05,
+      "loss": 3.9145,
+      "step": 2632
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1630181233870219e-05,
+      "loss": 4.4846,
+      "step": 2633
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1697724575838453e-05,
+      "loss": 4.8249,
+      "step": 2634
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1765451300597563e-05,
+      "loss": 4.3604,
+      "step": 2635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1833361267495257e-05,
+      "loss": 3.6993,
+      "step": 2636
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1901454335498741e-05,
+      "loss": 6.0283,
+      "step": 2637
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.1969730363194875e-05,
+      "loss": 5.215,
+      "step": 2638
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2038189208790695e-05,
+      "loss": 5.3595,
+      "step": 2639
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.210683073011345e-05,
+      "loss": 4.4039,
+      "step": 2640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2175654784611034e-05,
+      "loss": 5.093,
+      "step": 2641
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2244661229352295e-05,
+      "loss": 3.5903,
+      "step": 2642
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2313849921027287e-05,
+      "loss": 5.5689,
+      "step": 2643
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2383220715947585e-05,
+      "loss": 3.1709,
+      "step": 2644
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2452773470046531e-05,
+      "loss": 3.1529,
+      "step": 2645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2522508038879687e-05,
+      "loss": 6.0319,
+      "step": 2646
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2592424277624937e-05,
+      "loss": 5.2714,
+      "step": 2647
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2662522041082914e-05,
+      "loss": 3.8834,
+      "step": 2648
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2732801183677235e-05,
+      "loss": 4.8519,
+      "step": 2649
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2803261559454894e-05,
+      "loss": 4.9854,
+      "step": 2650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2873903022086476e-05,
+      "loss": 4.4817,
+      "step": 2651
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.2944725424866478e-05,
+      "loss": 4.9012,
+      "step": 2652
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3015728620713619e-05,
+      "loss": 3.9142,
+      "step": 2653
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.308691246217123e-05,
+      "loss": 2.7767,
+      "step": 2654
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3158276801407399e-05,
+      "loss": 4.6861,
+      "step": 2655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3229821490215411e-05,
+      "loss": 4.547,
+      "step": 2656
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3301546380013951e-05,
+      "loss": 3.3109,
+      "step": 2657
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.337345132184753e-05,
+      "loss": 3.6587,
+      "step": 2658
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.344553616638672e-05,
+      "loss": 3.0521,
+      "step": 2659
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3517800763928445e-05,
+      "loss": 6.1221,
+      "step": 2660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.3590244964396392e-05,
+      "loss": 5.0959,
+      "step": 2661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3662868617341195e-05,
+      "loss": 3.823,
+      "step": 2662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3735671571940822e-05,
+      "loss": 3.4493,
+      "step": 2663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3808653677000882e-05,
+      "loss": 5.8427,
+      "step": 2664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3881814780954906e-05,
+      "loss": 3.5039,
+      "step": 2665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.3955154731864761e-05,
+      "loss": 4.1194,
+      "step": 2666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.402867337742081e-05,
+      "loss": 4.0594,
+      "step": 2667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4102370564942336e-05,
+      "loss": 4.3601,
+      "step": 2668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4176246141377824e-05,
+      "loss": 5.1374,
+      "step": 2669
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4250299953305335e-05,
+      "loss": 4.3017,
+      "step": 2670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4324531846932731e-05,
+      "loss": 4.9115,
+      "step": 2671
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4398941668098065e-05,
+      "loss": 6.9005,
+      "step": 2672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4473529262269824e-05,
+      "loss": 4.7697,
+      "step": 2673
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4548294474547364e-05,
+      "loss": 5.7178,
+      "step": 2674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.462323714966115e-05,
+      "loss": 4.938,
+      "step": 2675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4698357131973084e-05,
+      "loss": 3.2088,
+      "step": 2676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.4773654265476888e-05,
+      "loss": 5.0267,
+      "step": 2677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.484912839379834e-05,
+      "loss": 3.8619,
+      "step": 2678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.492477936019565e-05,
+      "loss": 3.3877,
+      "step": 2679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5000607007559787e-05,
+      "loss": 4.3408,
+      "step": 2680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5076611178414757e-05,
+      "loss": 3.6659,
+      "step": 2681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5152791714918068e-05,
+      "loss": 4.3551,
+      "step": 2682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5229148458860854e-05,
+      "loss": 4.6482,
+      "step": 2683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.530568125166836e-05,
+      "loss": 4.9391,
+      "step": 2684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5382389934400164e-05,
+      "loss": 4.3198,
+      "step": 2685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5459274347750662e-05,
+      "loss": 4.5223,
+      "step": 2686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.553633433204923e-05,
+      "loss": 4.8133,
+      "step": 2687
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5613569727260656e-05,
+      "loss": 5.5434,
+      "step": 2688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5690980372985366e-05,
+      "loss": 4.428,
+      "step": 2689
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.5768566108459927e-05,
+      "loss": 5.0173,
+      "step": 2690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.584632677255723e-05,
+      "loss": 3.4918,
+      "step": 2691
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.59242622037869e-05,
+      "loss": 4.9854,
+      "step": 2692
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.600237224029564e-05,
+      "loss": 4.3795,
+      "step": 2693
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.6080656719867494e-05,
+      "loss": 3.3796,
+      "step": 2694
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.6159115479924247e-05,
+      "loss": 6.4341,
+      "step": 2695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.6237748357525716e-05,
+      "loss": 3.6448,
+      "step": 2696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.631655518937021e-05,
+      "loss": 3.4515,
+      "step": 2697
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.639553581179467e-05,
+      "loss": 4.5268,
+      "step": 2698
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.6474690060775166e-05,
+      "loss": 3.6097,
+      "step": 2699
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.6554017771927177e-05,
+      "loss": 3.6938,
+      "step": 2700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.6633518780505918e-05,
+      "loss": 5.61,
+      "step": 2701
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.6713192921406796e-05,
+      "loss": 3.6774,
+      "step": 2702
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.6793040029165553e-05,
+      "loss": 3.4431,
+      "step": 2703
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.68730599379588e-05,
+      "loss": 4.8614,
+      "step": 2704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.695325248160421e-05,
+      "loss": 4.8975,
+      "step": 2705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.7033617493560995e-05,
+      "loss": 4.2484,
+      "step": 2706
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.711415480693015e-05,
+      "loss": 4.7628,
+      "step": 2707
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.7194864254454855e-05,
+      "loss": 4.1564,
+      "step": 2708
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.727574566852087e-05,
+      "loss": 5.2961,
+      "step": 2709
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.735679888115671e-05,
+      "loss": 5.1348,
+      "step": 2710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.7438023724034193e-05,
+      "loss": 4.7555,
+      "step": 2711
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.7519420028468637e-05,
+      "loss": 4.529,
+      "step": 2712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.760098762541934e-05,
+      "loss": 3.7175,
+      "step": 2713
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.7682726345489818e-05,
+      "loss": 4.7574,
+      "step": 2714
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.7764636018928236e-05,
+      "loss": 3.8557,
+      "step": 2715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.78467164756277e-05,
+      "loss": 3.4553,
+      "step": 2716
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.792896754512663e-05,
+      "loss": 4.9963,
+      "step": 2717
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8011389056609206e-05,
+      "loss": 3.1596,
+      "step": 2718
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8093980838905543e-05,
+      "loss": 6.0232,
+      "step": 2719
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8176742720492225e-05,
+      "loss": 3.1655,
+      "step": 2720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.825967452949249e-05,
+      "loss": 4.9278,
+      "step": 2721
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.834277609367673e-05,
+      "loss": 3.1627,
+      "step": 2722
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8426047240462785e-05,
+      "loss": 4.844,
+      "step": 2723
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.850948779691638e-05,
+      "loss": 5.5447,
+      "step": 2724
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8593097589751306e-05,
+      "loss": 3.4611,
+      "step": 2725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.867687644532996e-05,
+      "loss": 5.3425,
+      "step": 2726
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8760824189663597e-05,
+      "loss": 4.2007,
+      "step": 2727
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8844940648412746e-05,
+      "loss": 5.7986,
+      "step": 2728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8929225646887607e-05,
+      "loss": 7.3336,
+      "step": 2729
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901367901004828e-05,
+      "loss": 5.1995,
+      "step": 2730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9098300562505246e-05,
+      "loss": 5.4587,
+      "step": 2731
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9183090128519688e-05,
+      "loss": 4.4974,
+      "step": 2732
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926804753200393e-05,
+      "loss": 6.3014,
+      "step": 2733
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9353172596521628e-05,
+      "loss": 3.442,
+      "step": 2734
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9438465145288344e-05,
+      "loss": 3.0177,
+      "step": 2735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.952392500117176e-05,
+      "loss": 4.978,
+      "step": 2736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9609551986692077e-05,
+      "loss": 5.1919,
+      "step": 2737
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.969534592402248e-05,
+      "loss": 5.6356,
+      "step": 2738
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9781306634989348e-05,
+      "loss": 2.9329,
+      "step": 2739
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9867433941072845e-05,
+      "loss": 4.5762,
+      "step": 2740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9953727663407018e-05,
+      "loss": 3.0756,
+      "step": 2741
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.004018762278038e-05,
+      "loss": 5.0733,
+      "step": 2742
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.012681363963619e-05,
+      "loss": 4.9086,
+      "step": 2743
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0213605534072823e-05,
+      "loss": 5.2056,
+      "step": 2744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.030056312584423e-05,
+      "loss": 5.0743,
+      "step": 2745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0387686234360182e-05,
+      "loss": 5.4814,
+      "step": 2746
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0474974678686753e-05,
+      "loss": 4.6479,
+      "step": 2747
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0562428277546596e-05,
+      "loss": 5.4262,
+      "step": 2748
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.065004684931947e-05,
+      "loss": 3.153,
+      "step": 2749
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.0737830212042462e-05,
+      "loss": 4.0419,
+      "step": 2750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.082577818341044e-05,
+      "loss": 4.8373,
+      "step": 2751
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.091389058077643e-05,
+      "loss": 3.8078,
+      "step": 2752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1002167221151924e-05,
+      "loss": 4.7557,
+      "step": 2753
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1090607921207406e-05,
+      "loss": 4.426,
+      "step": 2754
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.117921249727257e-05,
+      "loss": 3.9263,
+      "step": 2755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1267980765336883e-05,
+      "loss": 3.3324,
+      "step": 2756
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1356912541049757e-05,
+      "loss": 5.4076,
+      "step": 2757
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1446007639721088e-05,
+      "loss": 6.359,
+      "step": 2758
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1535265876321566e-05,
+      "loss": 6.196,
+      "step": 2759
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.162468706548313e-05,
+      "loss": 3.6725,
+      "step": 2760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1714271021499255e-05,
+      "loss": 4.7344,
+      "step": 2761
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1804017558325396e-05,
+      "loss": 5.6576,
+      "step": 2762
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1893926489579387e-05,
+      "loss": 5.2475,
+      "step": 2763
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.1983997628541764e-05,
+      "loss": 2.8385,
+      "step": 2764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.207423078815628e-05,
+      "loss": 5.1687,
+      "step": 2765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2164625781030124e-05,
+      "loss": 4.6748,
+      "step": 2766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.225518241943445e-05,
+      "loss": 3.6619,
+      "step": 2767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2345900515304707e-05,
+      "loss": 4.8146,
+      "step": 2768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2436779880240966e-05,
+      "loss": 5.0845,
+      "step": 2769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2527820325508452e-05,
+      "loss": 5.1862,
+      "step": 2770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2619021662037844e-05,
+      "loss": 7.3878,
+      "step": 2771
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.271038370042572e-05,
+      "loss": 3.3411,
+      "step": 2772
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.280190625093489e-05,
+      "loss": 4.9589,
+      "step": 2773
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2893589123494795e-05,
+      "loss": 4.4502,
+      "step": 2774
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.2985432127701932e-05,
+      "loss": 4.3463,
+      "step": 2775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.307743507282032e-05,
+      "loss": 5.2751,
+      "step": 2776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3169597767781727e-05,
+      "loss": 6.1464,
+      "step": 2777
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.326192002118619e-05,
+      "loss": 4.093,
+      "step": 2778
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.335440164130237e-05,
+      "loss": 4.6168,
+      "step": 2779
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3447042436067966e-05,
+      "loss": 5.5371,
+      "step": 2780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3539842213090158e-05,
+      "loss": 6.1278,
+      "step": 2781
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.363280077964589e-05,
+      "loss": 5.2483,
+      "step": 2782
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.3725917942682373e-05,
+      "loss": 4.3019,
+      "step": 2783
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.381919350881746e-05,
+      "loss": 5.8755,
+      "step": 2784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.391262728433996e-05,
+      "loss": 5.8125,
+      "step": 2785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4006219075210223e-05,
+      "loss": 5.4702,
+      "step": 2786
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.409996868706035e-05,
+      "loss": 5.3496,
+      "step": 2787
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4193875925194787e-05,
+      "loss": 5.91,
+      "step": 2788
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4287940594590554e-05,
+      "loss": 5.1826,
+      "step": 2789
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4382162499897732e-05,
+      "loss": 4.4538,
+      "step": 2790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.447654144543984e-05,
+      "loss": 4.9761,
+      "step": 2791
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4571077235214358e-05,
+      "loss": 3.8459,
+      "step": 2792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4665769672892923e-05,
+      "loss": 5.5397,
+      "step": 2793
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4760618561821925e-05,
+      "loss": 5.3467,
+      "step": 2794
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4855623705022767e-05,
+      "loss": 6.3503,
+      "step": 2795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.495078490519247e-05,
+      "loss": 5.0172,
+      "step": 2796
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.504610196470385e-05,
+      "loss": 4.1401,
+      "step": 2797
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5141574685606085e-05,
+      "loss": 4.716,
+      "step": 2798
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5237202869625076e-05,
+      "loss": 4.5514,
+      "step": 2799
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.53329863181639e-05,
+      "loss": 4.8127,
+      "step": 2800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.542892483230309e-05,
+      "loss": 4.4593,
+      "step": 2801
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5525018212801198e-05,
+      "loss": 6.0825,
+      "step": 2802
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5621266260095244e-05,
+      "loss": 4.3519,
+      "step": 2803
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5717668774300907e-05,
+      "loss": 4.5168,
+      "step": 2804
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.5814225555213145e-05,
+      "loss": 5.2397,
+      "step": 2805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.591093640230652e-05,
+      "loss": 5.692,
+      "step": 2806
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6007801114735608e-05,
+      "loss": 5.006,
+      "step": 2807
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6104819491335552e-05,
+      "loss": 4.0682,
+      "step": 2808
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6201991330622245e-05,
+      "loss": 4.8584,
+      "step": 2809
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6299316430792942e-05,
+      "loss": 5.6274,
+      "step": 2810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.639679458972655e-05,
+      "loss": 6.1195,
+      "step": 2811
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.649442560498422e-05,
+      "loss": 4.8906,
+      "step": 2812
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6592209273809564e-05,
+      "loss": 5.0084,
+      "step": 2813
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6690145393129194e-05,
+      "loss": 5.2282,
+      "step": 2814
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6788233759553115e-05,
+      "loss": 5.2282,
+      "step": 2815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.6886474169375198e-05,
+      "loss": 5.3123,
+      "step": 2816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.698486641857344e-05,
+      "loss": 4.1321,
+      "step": 2817
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7083410302810585e-05,
+      "loss": 4.1876,
+      "step": 2818
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.718210561743453e-05,
+      "loss": 4.51,
+      "step": 2819
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7280952157478567e-05,
+      "loss": 5.3088,
+      "step": 2820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7379949717662003e-05,
+      "loss": 5.2292,
+      "step": 2821
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7479098092390465e-05,
+      "loss": 3.2687,
+      "step": 2822
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.757839707575638e-05,
+      "loss": 4.6632,
+      "step": 2823
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.767784646153948e-05,
+      "loss": 5.2418,
+      "step": 2824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7777446043207034e-05,
+      "loss": 3.1718,
+      "step": 2825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.787719561391445e-05,
+      "loss": 5.3353,
+      "step": 2826
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.7977094966505613e-05,
+      "loss": 3.7578,
+      "step": 2827
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8077143893513393e-05,
+      "loss": 4.0663,
+      "step": 2828
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8177342187160004e-05,
+      "loss": 5.1683,
+      "step": 2829
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.827768963935744e-05,
+      "loss": 4.4744,
+      "step": 2830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8378186041707923e-05,
+      "loss": 4.4319,
+      "step": 2831
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8478831185504495e-05,
+      "loss": 3.6709,
+      "step": 2832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.857962486173099e-05,
+      "loss": 5.0612,
+      "step": 2833
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.868056686106302e-05,
+      "loss": 5.3213,
+      "step": 2834
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.878165697386813e-05,
+      "loss": 3.2847,
+      "step": 2835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8882894990206223e-05,
+      "loss": 3.7095,
+      "step": 2836
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.8984280699830047e-05,
+      "loss": 5.0416,
+      "step": 2837
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9085813892185608e-05,
+      "loss": 3.8525,
+      "step": 2838
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9187494356412716e-05,
+      "loss": 4.9079,
+      "step": 2839
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9289321881345234e-05,
+      "loss": 5.217,
+      "step": 2840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.939129625551166e-05,
+      "loss": 4.8186,
+      "step": 2841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9493417267135494e-05,
+      "loss": 4.2484,
+      "step": 2842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9595684704135695e-05,
+      "loss": 3.913,
+      "step": 2843
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9698098354127234e-05,
+      "loss": 3.9944,
+      "step": 2844
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.980065800442131e-05,
+      "loss": 4.7216,
+      "step": 2845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9903363442025978e-05,
+      "loss": 3.1855,
+      "step": 2846
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.0006214453646465e-05,
+      "loss": 4.8023,
+      "step": 2847
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.0109210825685786e-05,
+      "loss": 4.1927,
+      "step": 2848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.0212352344245043e-05,
+      "loss": 3.2494,
+      "step": 2849
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.0315638795123714e-05,
+      "loss": 5.0348,
+      "step": 2850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.041906996382057e-05,
+      "loss": 3.46,
+      "step": 2851
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.05226456355337e-05,
+      "loss": 4.8573,
+      "step": 2852
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.062636559516108e-05,
+      "loss": 2.9765,
+      "step": 2853
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.0730229627301046e-05,
+      "loss": 3.9591,
+      "step": 2854
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.0834237516252797e-05,
+      "loss": 4.568,
+      "step": 2855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.0938389046016717e-05,
+      "loss": 3.7744,
+      "step": 2856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.104268400029485e-05,
+      "loss": 3.7767,
+      "step": 2857
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.114712216249144e-05,
+      "loss": 3.9868,
+      "step": 2858
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.125170331571332e-05,
+      "loss": 6.3082,
+      "step": 2859
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.135642724277034e-05,
+      "loss": 4.6875,
+      "step": 2860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.1461293726175844e-05,
+      "loss": 5.7008,
+      "step": 2861
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.156630254814713e-05,
+      "loss": 5.6167,
+      "step": 2862
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.167145349060586e-05,
+      "loss": 4.7071,
+      "step": 2863
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.177674633517863e-05,
+      "loss": 4.5136,
+      "step": 2864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.18821808631973e-05,
+      "loss": 5.1025,
+      "step": 2865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.198775685569938e-05,
+      "loss": 5.5166,
+      "step": 2866
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.209347409342875e-05,
+      "loss": 5.4513,
+      "step": 2867
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.219933235683587e-05,
+      "loss": 6.8251,
+      "step": 2868
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.2305331426078344e-05,
+      "loss": 5.708,
+      "step": 2869
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.241147108102134e-05,
+      "loss": 5.1879,
+      "step": 2870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.2517751101238125e-05,
+      "loss": 5.1916,
+      "step": 2871
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.2624171266010384e-05,
+      "loss": 5.6021,
+      "step": 2872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.27307313543288e-05,
+      "loss": 3.6728,
+      "step": 2873
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.283743114489342e-05,
+      "loss": 4.5463,
+      "step": 2874
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.294427041611424e-05,
+      "loss": 5.6124,
+      "step": 2875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.3051248946111534e-05,
+      "loss": 5.3076,
+      "step": 2876
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.315836651271638e-05,
+      "loss": 5.9898,
+      "step": 2877
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.3265622893471084e-05,
+      "loss": 5.3688,
+      "step": 2878
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.33730178656297e-05,
+      "loss": 5.0493,
+      "step": 2879
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.348055120615849e-05,
+      "loss": 5.0667,
+      "step": 2880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.3588222691736326e-05,
+      "loss": 3.4357,
+      "step": 2881
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.369603209875513e-05,
+      "loss": 4.7271,
+      "step": 2882
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.380397920332046e-05,
+      "loss": 6.3724,
+      "step": 2883
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.39120637812519e-05,
+      "loss": 4.8541,
+      "step": 2884
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.4020285608083546e-05,
+      "loss": 5.2854,
+      "step": 2885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.41286444590644e-05,
+      "loss": 3.8992,
+      "step": 2886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.423714010915901e-05,
+      "loss": 5.5549,
+      "step": 2887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.43457723330477e-05,
+      "loss": 4.6999,
+      "step": 2888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.4454540905127254e-05,
+      "loss": 5.2352,
+      "step": 2889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.45634455995112e-05,
+      "loss": 4.8541,
+      "step": 2890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.467248619003055e-05,
+      "loss": 3.7078,
+      "step": 2891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.47816624502338e-05,
+      "loss": 3.0023,
+      "step": 2892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.489097415338802e-05,
+      "loss": 5.4215,
+      "step": 2893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.500042107247878e-05,
+      "loss": 4.7377,
+      "step": 2894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.511000298021092e-05,
+      "loss": 4.7423,
+      "step": 2895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.5219719649008905e-05,
+      "loss": 4.3186,
+      "step": 2896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.5329570851017423e-05,
+      "loss": 4.1556,
+      "step": 2897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.5439556358101554e-05,
+      "loss": 5.0905,
+      "step": 2898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.554967594184761e-05,
+      "loss": 5.8162,
+      "step": 2899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.565992937356356e-05,
+      "loss": 3.4115,
+      "step": 2900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.577031642427926e-05,
+      "loss": 3.9174,
+      "step": 2901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.588083686474708e-05,
+      "loss": 5.0624,
+      "step": 2902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.599149046544239e-05,
+      "loss": 2.9312,
+      "step": 2903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.610227699656403e-05,
+      "loss": 4.8182,
+      "step": 2904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.621319622803477e-05,
+      "loss": 4.9058,
+      "step": 2905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.6324247929501785e-05,
+      "loss": 7.023,
+      "step": 2906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.643543187033713e-05,
+      "loss": 4.0015,
+      "step": 2907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.65467478196382e-05,
+      "loss": 5.0523,
+      "step": 2908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.6658195546228414e-05,
+      "loss": 5.2518,
+      "step": 2909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.676977481865732e-05,
+      "loss": 5.1088,
+      "step": 2910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.688148540520135e-05,
+      "loss": 4.9215,
+      "step": 2911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.6993327073864245e-05,
+      "loss": 4.1216,
+      "step": 2912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.710529959237754e-05,
+      "loss": 3.568,
+      "step": 2913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.721740272820088e-05,
+      "loss": 4.2404,
+      "step": 2914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.7329636248522734e-05,
+      "loss": 5.7644,
+      "step": 2915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.744199992026096e-05,
+      "loss": 5.3534,
+      "step": 2916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.755449351006293e-05,
+      "loss": 4.5897,
+      "step": 2917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.766711678430622e-05,
+      "loss": 4.4016,
+      "step": 2918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.777986950909914e-05,
+      "loss": 5.8173,
+      "step": 2919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.789275145028115e-05,
+      "loss": 4.357,
+      "step": 2920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.800576237342333e-05,
+      "loss": 6.2895,
+      "step": 2921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.811890204382891e-05,
+      "loss": 5.2003,
+      "step": 2922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.823217022653377e-05,
+      "loss": 3.1866,
+      "step": 2923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.8345566686306787e-05,
+      "loss": 5.4916,
+      "step": 2924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.845909118765068e-05,
+      "loss": 4.3146,
+      "step": 2925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.857274349480202e-05,
+      "loss": 3.7792,
+      "step": 2926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.868652337173205e-05,
+      "loss": 4.6089,
+      "step": 2927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.880043058214706e-05,
+      "loss": 2.7895,
+      "step": 2928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.8914464889489e-05,
+      "loss": 3.2553,
+      "step": 2929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.9028626056935615e-05,
+      "loss": 5.2156,
+      "step": 2930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.914291384740136e-05,
+      "loss": 4.2774,
+      "step": 2931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.925732802353788e-05,
+      "loss": 6.4811,
+      "step": 2932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.9371868347734076e-05,
+      "loss": 4.1033,
+      "step": 2933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.948653458211702e-05,
+      "loss": 4.2698,
+      "step": 2934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.960132648855224e-05,
+      "loss": 4.9654,
+      "step": 2935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.97162438286443e-05,
+      "loss": 4.6844,
+      "step": 2936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.983128636373724e-05,
+      "loss": 5.1565,
+      "step": 2937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.994645385491516e-05,
+      "loss": 4.5314,
+      "step": 2938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.006174606300257e-05,
+      "loss": 5.1679,
+      "step": 2939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.017716274856497e-05,
+      "loss": 3.6,
+      "step": 2940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.0292703671909546e-05,
+      "loss": 4.8621,
+      "step": 2941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.040836859308524e-05,
+      "loss": 2.7467,
+      "step": 2942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.052415727188358e-05,
+      "loss": 3.7405,
+      "step": 2943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.064006946783904e-05,
+      "loss": 4.6909,
+      "step": 2944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.075610494022969e-05,
+      "loss": 4.8273,
+      "step": 2945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.087226344807729e-05,
+      "loss": 4.4348,
+      "step": 2946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.0988544750148474e-05,
+      "loss": 3.7075,
+      "step": 2947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.110494860495463e-05,
+      "loss": 4.771,
+      "step": 2948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.122147477075268e-05,
+      "loss": 4.8523,
+      "step": 2949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.1338123005545524e-05,
+      "loss": 4.8571,
+      "step": 2950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.145489306708258e-05,
+      "loss": 5.8802,
+      "step": 2951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.157178471286023e-05,
+      "loss": 3.8524,
+      "step": 2952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.168879770012239e-05,
+      "loss": 3.8701,
+      "step": 2953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.180593178586095e-05,
+      "loss": 4.8611,
+      "step": 2954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.192318672681625e-05,
+      "loss": 4.3735,
+      "step": 2955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2040562279477845e-05,
+      "loss": 4.7176,
+      "step": 2956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2158058200084616e-05,
+      "loss": 4.9615,
+      "step": 2957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2275674244625515e-05,
+      "loss": 5.465,
+      "step": 2958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.239341016884003e-05,
+      "loss": 4.7815,
+      "step": 2959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2511265728218706e-05,
+      "loss": 4.3946,
+      "step": 2960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.262924067800369e-05,
+      "loss": 5.267,
+      "step": 2961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.2747334773188896e-05,
+      "loss": 3.6612,
+      "step": 2962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.286554776852123e-05,
+      "loss": 4.9898,
+      "step": 2963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.298387941850041e-05,
+      "loss": 3.5918,
+      "step": 2964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.310232947737977e-05,
+      "loss": 6.7637,
+      "step": 2965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.3220897699166755e-05,
+      "loss": 4.2481,
+      "step": 2966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.333958383762342e-05,
+      "loss": 4.9884,
+      "step": 2967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.3458387646266916e-05,
+      "loss": 5.405,
+      "step": 2968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.357730887837006e-05,
+      "loss": 5.051,
+      "step": 2969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.369634728696176e-05,
+      "loss": 4.1485,
+      "step": 2970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.3815502624827534e-05,
+      "loss": 5.5508,
+      "step": 2971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.393477464451027e-05,
+      "loss": 4.6881,
+      "step": 2972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.405416309831032e-05,
+      "loss": 6.3557,
+      "step": 2973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4173667738286314e-05,
+      "loss": 5.5161,
+      "step": 2974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4293288316255586e-05,
+      "loss": 4.966,
+      "step": 2975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.441302458379469e-05,
+      "loss": 3.4671,
+      "step": 2976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.453287629223999e-05,
+      "loss": 4.0131,
+      "step": 2977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.465284319268782e-05,
+      "loss": 5.7163,
+      "step": 2978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.477292503599573e-05,
+      "loss": 4.993,
+      "step": 2979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.489312157278225e-05,
+      "loss": 2.81,
+      "step": 2980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.501343255342781e-05,
+      "loss": 4.0606,
+      "step": 2981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.513385772807518e-05,
+      "loss": 4.7937,
+      "step": 2982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.525439684662996e-05,
+      "loss": 2.9382,
+      "step": 2983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.537504965876108e-05,
+      "loss": 6.3382,
+      "step": 2984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.549581591390143e-05,
+      "loss": 4.9468,
+      "step": 2985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.561669536124825e-05,
+      "loss": 4.31,
+      "step": 2986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5737687749763646e-05,
+      "loss": 4.4991,
+      "step": 2987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.585879282817541e-05,
+      "loss": 3.6054,
+      "step": 2988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.598001034497706e-05,
+      "loss": 4.1567,
+      "step": 2989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.61013400484287e-05,
+      "loss": 3.9139,
+      "step": 2990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6222781686557414e-05,
+      "loss": 4.975,
+      "step": 2991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.634433500715788e-05,
+      "loss": 5.3853,
+      "step": 2992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6465999757792865e-05,
+      "loss": 4.6138,
+      "step": 2993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.658777568579344e-05,
+      "loss": 5.1055,
+      "step": 2994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.6709662538260246e-05,
+      "loss": 4.2404,
+      "step": 2995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.683166006206324e-05,
+      "loss": 4.206,
+      "step": 2996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.695376800384263e-05,
+      "loss": 2.9039,
+      "step": 2997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7075986110009315e-05,
+      "loss": 4.7362,
+      "step": 2998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.719831412674541e-05,
+      "loss": 4.3054,
+      "step": 2999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7320751800004745e-05,
+      "loss": 5.4742,
+      "step": 3000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.744329887551344e-05,
+      "loss": 4.3044,
+      "step": 3001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.756595509877044e-05,
+      "loss": 3.6108,
+      "step": 3002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.768872021504788e-05,
+      "loss": 4.7127,
+      "step": 3003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.781159396939203e-05,
+      "loss": 3.9732,
+      "step": 3004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.793457610662329e-05,
+      "loss": 4.9086,
+      "step": 3005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.805766637133705e-05,
+      "loss": 4.4656,
+      "step": 3006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8180864507904164e-05,
+      "loss": 3.8541,
+      "step": 3007
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.830417026047145e-05,
+      "loss": 4.9191,
+      "step": 3008
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8427583372962305e-05,
+      "loss": 5.1332,
+      "step": 3009
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8551103589076874e-05,
+      "loss": 4.3765,
+      "step": 3010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.86747306522933e-05,
+      "loss": 4.9415,
+      "step": 3011
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8798464305867574e-05,
+      "loss": 4.1015,
+      "step": 3012
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8922304292834355e-05,
+      "loss": 4.6934,
+      "step": 3013
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9046250356007495e-05,
+      "loss": 4.44,
+      "step": 3014
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9170302237980545e-05,
+      "loss": 5.2599,
+      "step": 3015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.92944596811273e-05,
+      "loss": 5.8429,
+      "step": 3016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.941872242760234e-05,
+      "loss": 5.4617,
+      "step": 3017
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9543090219341434e-05,
+      "loss": 5.4294,
+      "step": 3018
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.96675627980625e-05,
+      "loss": 5.5098,
+      "step": 3019
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.979213990526552e-05,
+      "loss": 5.2452,
+      "step": 3020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.991682128223356e-05,
+      "loss": 4.387,
+      "step": 3021
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.004160667003308e-05,
+      "loss": 4.6348,
+      "step": 3022
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.016649580951456e-05,
+      "loss": 3.9918,
+      "step": 3023
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.029148844131301e-05,
+      "loss": 6.3382,
+      "step": 3024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.0416584305848504e-05,
+      "loss": 4.4892,
+      "step": 3025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.0541783143326734e-05,
+      "loss": 5.8703,
+      "step": 3026
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.066708469373955e-05,
+      "loss": 4.5608,
+      "step": 3027
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.079248869686546e-05,
+      "loss": 4.4577,
+      "step": 3028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.0917994892270246e-05,
+      "loss": 4.3769,
+      "step": 3029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.104360301930742e-05,
+      "loss": 3.6293,
+      "step": 3030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.116931281711883e-05,
+      "loss": 4.3392,
+      "step": 3031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.129512402463517e-05,
+      "loss": 2.8716,
+      "step": 3032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.142103638057654e-05,
+      "loss": 5.8079,
+      "step": 3033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.154704962345288e-05,
+      "loss": 4.2053,
+      "step": 3034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.167316349156488e-05,
+      "loss": 3.4079,
+      "step": 3035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.179937772300398e-05,
+      "loss": 5.6544,
+      "step": 3036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.19256920556533e-05,
+      "loss": 3.8624,
+      "step": 3037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.205210622718806e-05,
+      "loss": 3.755,
+      "step": 3038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.217861997507613e-05,
+      "loss": 3.6521,
+      "step": 3039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.2305233036578595e-05,
+      "loss": 3.7974,
+      "step": 3040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.243194514875031e-05,
+      "loss": 4.544,
+      "step": 3041
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.255875604844038e-05,
+      "loss": 3.9436,
+      "step": 3042
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.268566547229278e-05,
+      "loss": 5.2605,
+      "step": 3043
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.281267315674687e-05,
+      "loss": 6.0769,
+      "step": 3044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.293977883803794e-05,
+      "loss": 5.2817,
+      "step": 3045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.306698225219778e-05,
+      "loss": 5.2733,
+      "step": 3046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.31942831350552e-05,
+      "loss": 5.3048,
+      "step": 3047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.332168122223657e-05,
+      "loss": 4.7177,
+      "step": 3048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.3449176249166455e-05,
+      "loss": 5.648,
+      "step": 3049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.357676795106793e-05,
+      "loss": 5.506,
+      "step": 3050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.370445606296362e-05,
+      "loss": 5.9182,
+      "step": 3051
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.383224031967562e-05,
+      "loss": 5.4488,
+      "step": 3052
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.396012045582648e-05,
+      "loss": 2.6413,
+      "step": 3053
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.408809620583964e-05,
+      "loss": 3.7849,
+      "step": 3054
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.4216167303939946e-05,
+      "loss": 5.6333,
+      "step": 3055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.434433348415422e-05,
+      "loss": 4.1983,
+      "step": 3056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.447259448031186e-05,
+      "loss": 3.1909,
+      "step": 3057
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.4600950026045305e-05,
+      "loss": 4.2355,
+      "step": 3058
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.472939985479069e-05,
+      "loss": 4.8008,
+      "step": 3059
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.48579436997883e-05,
+      "loss": 4.5708,
+      "step": 3060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.4986581294083206e-05,
+      "loss": 4.9617,
+      "step": 3061
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.511531237052574e-05,
+      "loss": 4.7278,
+      "step": 3062
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.524413666177217e-05,
+      "loss": 4.9426,
+      "step": 3063
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.537305390028509e-05,
+      "loss": 4.8215,
+      "step": 3064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.550206381833411e-05,
+      "loss": 4.7272,
+      "step": 3065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.5631166147996304e-05,
+      "loss": 5.2908,
+      "step": 3066
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.5760360621157036e-05,
+      "loss": 4.583,
+      "step": 3067
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.5889646969510076e-05,
+      "loss": 5.5249,
+      "step": 3068
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.6019024924558484e-05,
+      "loss": 3.294,
+      "step": 3069
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.614849421761508e-05,
+      "loss": 3.9945,
+      "step": 3070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.627805457980295e-05,
+      "loss": 6.3494,
+      "step": 3071
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.6407705742056114e-05,
+      "loss": 4.3664,
+      "step": 3072
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.653744743511997e-05,
+      "loss": 4.2833,
+      "step": 3073
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.6667279389551965e-05,
+      "loss": 5.0319,
+      "step": 3074
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.679720133572204e-05,
+      "loss": 4.6775,
+      "step": 3075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.692721300381324e-05,
+      "loss": 4.5038,
+      "step": 3076
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.7057314123822324e-05,
+      "loss": 4.0808,
+      "step": 3077
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.7187504425560254e-05,
+      "loss": 2.8711,
+      "step": 3078
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.731778363865279e-05,
+      "loss": 3.3053,
+      "step": 3079
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.744815149254105e-05,
+      "loss": 4.5281,
+      "step": 3080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.757860771648196e-05,
+      "loss": 5.1124,
+      "step": 3081
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.770915203954916e-05,
+      "loss": 4.3766,
+      "step": 3082
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.7839784190633165e-05,
+      "loss": 3.9392,
+      "step": 3083
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.797050389844207e-05,
+      "loss": 3.6786,
+      "step": 3084
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.810131089150222e-05,
+      "loss": 4.0541,
+      "step": 3085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.82322048981586e-05,
+      "loss": 3.8526,
+      "step": 3086
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.836318564657559e-05,
+      "loss": 5.3217,
+      "step": 3087
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.849425286473734e-05,
+      "loss": 4.9292,
+      "step": 3088
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.862540628044848e-05,
+      "loss": 4.1485,
+      "step": 3089
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.8756645621334604e-05,
+      "loss": 3.8337,
+      "step": 3090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.888797061484288e-05,
+      "loss": 3.8537,
+      "step": 3091
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.901938098824255e-05,
+      "loss": 5.7526,
+      "step": 3092
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.915087646862559e-05,
+      "loss": 4.3798,
+      "step": 3093
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.928245678290723e-05,
+      "loss": 5.874,
+      "step": 3094
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.941412165782645e-05,
+      "loss": 4.06,
+      "step": 3095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.954587081994672e-05,
+      "loss": 3.8595,
+      "step": 3096
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.967770399565631e-05,
+      "loss": 4.4165,
+      "step": 3097
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.98096209111693e-05,
+      "loss": 2.6657,
+      "step": 3098
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.994162129252557e-05,
+      "loss": 5.3533,
+      "step": 3099
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.007370486559181e-05,
+      "loss": 5.4035,
+      "step": 3100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.020587135606186e-05,
+      "loss": 4.3067,
+      "step": 3101
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.033812048945744e-05,
+      "loss": 4.6152,
+      "step": 3102
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.0470451991128596e-05,
+      "loss": 4.9371,
+      "step": 3103
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.060286558625432e-05,
+      "loss": 3.958,
+      "step": 3104
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.073536099984312e-05,
+      "loss": 4.4186,
+      "step": 3105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.0867937956733555e-05,
+      "loss": 4.6201,
+      "step": 3106
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.100059618159493e-05,
+      "loss": 5.0163,
+      "step": 3107
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.113333539892765e-05,
+      "loss": 4.9004,
+      "step": 3108
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.126615533306402e-05,
+      "loss": 4.2404,
+      "step": 3109
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.139905570816867e-05,
+      "loss": 5.2373,
+      "step": 3110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.153203624823917e-05,
+      "loss": 4.0819,
+      "step": 3111
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.166509667710663e-05,
+      "loss": 3.5485,
+      "step": 3112
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.179823671843615e-05,
+      "loss": 3.6149,
+      "step": 3113
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.193145609572773e-05,
+      "loss": 4.6557,
+      "step": 3114
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.20647545323164e-05,
+      "loss": 4.5248,
+      "step": 3115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.219813175137304e-05,
+      "loss": 3.2923,
+      "step": 3116
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.233158747590496e-05,
+      "loss": 4.205,
+      "step": 3117
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.246512142875638e-05,
+      "loss": 4.5367,
+      "step": 3118
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.259873333260913e-05,
+      "loss": 4.1186,
+      "step": 3119
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.27324229099831e-05,
+      "loss": 4.98,
+      "step": 3120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.286618988323686e-05,
+      "loss": 4.2723,
+      "step": 3121
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.30000339745683e-05,
+      "loss": 4.3186,
+      "step": 3122
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.313395490601512e-05,
+      "loss": 4.0627,
+      "step": 3123
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.326795239945546e-05,
+      "loss": 5.0989,
+      "step": 3124
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.340202617660842e-05,
+      "loss": 5.7587,
+      "step": 3125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.353617595903472e-05,
+      "loss": 4.2458,
+      "step": 3126
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.367040146813721e-05,
+      "loss": 4.6295,
+      "step": 3127
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.38047024251615e-05,
+      "loss": 4.8352,
+      "step": 3128
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.393907855119639e-05,
+      "loss": 5.5297,
+      "step": 3129
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.407352956717486e-05,
+      "loss": 4.7036,
+      "step": 3130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.420805519387407e-05,
+      "loss": 3.674,
+      "step": 3131
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.434265515191638e-05,
+      "loss": 4.8487,
+      "step": 3132
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.447732916176971e-05,
+      "loss": 5.6632,
+      "step": 3133
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.461207694374823e-05,
+      "loss": 3.2009,
+      "step": 3134
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.474689821801291e-05,
+      "loss": 4.0208,
+      "step": 3135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.488179270457206e-05,
+      "loss": 6.1809,
+      "step": 3136
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.5016760123282e-05,
+      "loss": 5.7692,
+      "step": 3137
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.51518001938475e-05,
+      "loss": 4.2308,
+      "step": 3138
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.528691263582254e-05,
+      "loss": 4.1323,
+      "step": 3139
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.542209716861075e-05,
+      "loss": 6.3568,
+      "step": 3140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.555735351146607e-05,
+      "loss": 4.7737,
+      "step": 3141
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.56926813834933e-05,
+      "loss": 5.506,
+      "step": 3142
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.582808050364866e-05,
+      "loss": 5.6808,
+      "step": 3143
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.596355059074046e-05,
+      "loss": 2.771,
+      "step": 3144
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.609909136342948e-05,
+      "loss": 4.7224,
+      "step": 3145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.623470254023005e-05,
+      "loss": 3.323,
+      "step": 3146
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.637038383950993e-05,
+      "loss": 4.9104,
+      "step": 3147
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.650613497949141e-05,
+      "loss": 5.1601,
+      "step": 3148
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.664195567825169e-05,
+      "loss": 4.9453,
+      "step": 3149
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.677784565372353e-05,
+      "loss": 4.8145,
+      "step": 3150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.691380462369584e-05,
+      "loss": 3.7021,
+      "step": 3151
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.704983230581424e-05,
+      "loss": 3.3573,
+      "step": 3152
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.718592841758158e-05,
+      "loss": 4.2892,
+      "step": 3153
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.732209267635872e-05,
+      "loss": 4.4967,
+      "step": 3154
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.745832479936491e-05,
+      "loss": 6.3788,
+      "step": 3155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.759462450367848e-05,
+      "loss": 4.4432,
+      "step": 3156
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.77309915062374e-05,
+      "loss": 3.7922,
+      "step": 3157
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.786742552383994e-05,
+      "loss": 4.9319,
+      "step": 3158
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.800392627314506e-05,
+      "loss": 4.7077,
+      "step": 3159
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.814049347067323e-05,
+      "loss": 6.722,
+      "step": 3160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.827712683280703e-05,
+      "loss": 4.9851,
+      "step": 3161
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.841382607579143e-05,
+      "loss": 4.4019,
+      "step": 3162
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.855059091573468e-05,
+      "loss": 6.4544,
+      "step": 3163
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.868742106860879e-05,
+      "loss": 4.0566,
+      "step": 3164
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.882431625025011e-05,
+      "loss": 5.7235,
+      "step": 3165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.896127617635997e-05,
+      "loss": 6.4063,
+      "step": 3166
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.909830056250524e-05,
+      "loss": 5.2407,
+      "step": 3167
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.923538912411891e-05,
+      "loss": 4.6873,
+      "step": 3168
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.937254157650067e-05,
+      "loss": 2.7588,
+      "step": 3169
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.950975763481757e-05,
+      "loss": 3.8742,
+      "step": 3170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.964703701410456e-05,
+      "loss": 4.3269,
+      "step": 3171
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.978437942926502e-05,
+      "loss": 4.1852,
+      "step": 3172
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 6.992178459507151e-05,
+      "loss": 3.9385,
+      "step": 3173
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.005925222616624e-05,
+      "loss": 6.3996,
+      "step": 3174
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.019678203706166e-05,
+      "loss": 4.8433,
+      "step": 3175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.033437374214104e-05,
+      "loss": 3.1758,
+      "step": 3176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.047202705565937e-05,
+      "loss": 3.6665,
+      "step": 3177
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.060974169174341e-05,
+      "loss": 5.2282,
+      "step": 3178
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.074751736439267e-05,
+      "loss": 5.0501,
+      "step": 3179
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.088535378747989e-05,
+      "loss": 5.8634,
+      "step": 3180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.102325067475167e-05,
+      "loss": 4.5157,
+      "step": 3181
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.1161207739829e-05,
+      "loss": 4.1344,
+      "step": 3182
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.129922469620795e-05,
+      "loss": 4.7075,
+      "step": 3183
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.143730125726017e-05,
+      "loss": 3.8198,
+      "step": 3184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.157543713623352e-05,
+      "loss": 3.1285,
+      "step": 3185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.171363204625269e-05,
+      "loss": 3.1457,
+      "step": 3186
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.185188570031981e-05,
+      "loss": 3.3405,
+      "step": 3187
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.199019781131493e-05,
+      "loss": 4.0315,
+      "step": 3188
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.212856809199675e-05,
+      "loss": 5.1576,
+      "step": 3189
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.226699625500317e-05,
+      "loss": 3.1151,
+      "step": 3190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.240548201285188e-05,
+      "loss": 5.6132,
+      "step": 3191
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.254402507794085e-05,
+      "loss": 5.5107,
+      "step": 3192
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.26826251625493e-05,
+      "loss": 5.214,
+      "step": 3193
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.282128197883781e-05,
+      "loss": 5.3139,
+      "step": 3194
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.295999523884918e-05,
+      "loss": 2.4369,
+      "step": 3195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.309876465450905e-05,
+      "loss": 4.6223,
+      "step": 3196
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.323758993762635e-05,
+      "loss": 4.0648,
+      "step": 3197
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.337647079989408e-05,
+      "loss": 4.7291,
+      "step": 3198
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.351540695288974e-05,
+      "loss": 3.6318,
+      "step": 3199
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.365439810807606e-05,
+      "loss": 3.5697,
+      "step": 3200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.37934439768015e-05,
+      "loss": 4.8304,
+      "step": 3201
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.393254427030096e-05,
+      "loss": 5.8709,
+      "step": 3202
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.40716986996962e-05,
+      "loss": 4.589,
+      "step": 3203
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.421090697599669e-05,
+      "loss": 4.9908,
+      "step": 3204
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.435016881009995e-05,
+      "loss": 3.0172,
+      "step": 3205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.448948391279234e-05,
+      "loss": 4.189,
+      "step": 3206
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.462885199474958e-05,
+      "loss": 3.9279,
+      "step": 3207
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.476827276653728e-05,
+      "loss": 5.0091,
+      "step": 3208
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.490774593861187e-05,
+      "loss": 4.2665,
+      "step": 3209
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.504727122132074e-05,
+      "loss": 3.8986,
+      "step": 3210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.518684832490308e-05,
+      "loss": 4.0921,
+      "step": 3211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.532647695949052e-05,
+      "loss": 5.4428,
+      "step": 3212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.546615683510763e-05,
+      "loss": 5.5976,
+      "step": 3213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.560588766167258e-05,
+      "loss": 4.8782,
+      "step": 3214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.574566914899777e-05,
+      "loss": 5.9321,
+      "step": 3215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.588550100679022e-05,
+      "loss": 3.4927,
+      "step": 3216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.602538294465273e-05,
+      "loss": 4.5208,
+      "step": 3217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.616531467208362e-05,
+      "loss": 4.814,
+      "step": 3218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.630529589847808e-05,
+      "loss": 5.1078,
+      "step": 3219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.644532633312844e-05,
+      "loss": 2.4205,
+      "step": 3220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.658540568522487e-05,
+      "loss": 3.7707,
+      "step": 3221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.672553366385593e-05,
+      "loss": 5.2507,
+      "step": 3222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.686570997800906e-05,
+      "loss": 4.5105,
+      "step": 3223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.700593433657167e-05,
+      "loss": 5.8687,
+      "step": 3224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.714620644833106e-05,
+      "loss": 4.0311,
+      "step": 3225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.728652602197545e-05,
+      "loss": 4.2707,
+      "step": 3226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.742689276609457e-05,
+      "loss": 3.8342,
+      "step": 3227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.756730638918008e-05,
+      "loss": 4.5165,
+      "step": 3228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.770776659962636e-05,
+      "loss": 5.0164,
+      "step": 3229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.784827310573099e-05,
+      "loss": 5.392,
+      "step": 3230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.798882561569544e-05,
+      "loss": 4.1127,
+      "step": 3231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.812942383762554e-05,
+      "loss": 3.3233,
+      "step": 3232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.827006747953255e-05,
+      "loss": 3.6644,
+      "step": 3233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.841075624933285e-05,
+      "loss": 4.0838,
+      "step": 3234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.855148985484947e-05,
+      "loss": 5.7169,
+      "step": 3235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.869226800381226e-05,
+      "loss": 5.0934,
+      "step": 3236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.883309040385851e-05,
+      "loss": 4.6129,
+      "step": 3237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.897395676253365e-05,
+      "loss": 4.1147,
+      "step": 3238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.911486678729172e-05,
+      "loss": 3.7758,
+      "step": 3239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.925582018549637e-05,
+      "loss": 5.1628,
+      "step": 3240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.939681666442085e-05,
+      "loss": 3.4766,
+      "step": 3241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.953785593124906e-05,
+      "loss": 4.0748,
+      "step": 3242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.967893769307606e-05,
+      "loss": 5.994,
+      "step": 3243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.982006165690861e-05,
+      "loss": 4.7607,
+      "step": 3244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.99612275296659e-05,
+      "loss": 5.3195,
+      "step": 3245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.010243501818001e-05,
+      "loss": 4.3282,
+      "step": 3246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.024368382919662e-05,
+      "loss": 5.7524,
+      "step": 3247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.038497366937554e-05,
+      "loss": 4.0921,
+      "step": 3248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.052630424529165e-05,
+      "loss": 4.0027,
+      "step": 3249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.066767526343479e-05,
+      "loss": 3.6497,
+      "step": 3250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.080908643021108e-05,
+      "loss": 5.9963,
+      "step": 3251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.095053745194325e-05,
+      "loss": 4.7584,
+      "step": 3252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.10920280348712e-05,
+      "loss": 4.1871,
+      "step": 3253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.12335578851527e-05,
+      "loss": 4.1476,
+      "step": 3254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.13751267088639e-05,
+      "loss": 3.6327,
+      "step": 3255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.151673421200022e-05,
+      "loss": 4.9487,
+      "step": 3256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.165838010047655e-05,
+      "loss": 2.8864,
+      "step": 3257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.18000640801281e-05,
+      "loss": 4.6473,
+      "step": 3258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.1941785856711e-05,
+      "loss": 4.9501,
+      "step": 3259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.208354513590288e-05,
+      "loss": 4.9591,
+      "step": 3260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.222534162330349e-05,
+      "loss": 5.9733,
+      "step": 3261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.236717502443533e-05,
+      "loss": 4.2254,
+      "step": 3262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.250904504474417e-05,
+      "loss": 4.574,
+      "step": 3263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.26509513895997e-05,
+      "loss": 4.3298,
+      "step": 3264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.279289376429653e-05,
+      "loss": 4.6828,
+      "step": 3265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.293487187405387e-05,
+      "loss": 4.8897,
+      "step": 3266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.307688542401712e-05,
+      "loss": 4.4775,
+      "step": 3267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.321893411925794e-05,
+      "loss": 4.0772,
+      "step": 3268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.336101766477506e-05,
+      "loss": 4.3512,
+      "step": 3269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.350313576549477e-05,
+      "loss": 3.7682,
+      "step": 3270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.364528812627157e-05,
+      "loss": 4.0285,
+      "step": 3271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.378747445188902e-05,
+      "loss": 4.9946,
+      "step": 3272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.392969444705993e-05,
+      "loss": 4.3656,
+      "step": 3273
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.407194781642723e-05,
+      "loss": 5.9011,
+      "step": 3274
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.421423426456458e-05,
+      "loss": 5.6382,
+      "step": 3275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.435655349597689e-05,
+      "loss": 5.3843,
+      "step": 3276
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.449890521510107e-05,
+      "loss": 6.1603,
+      "step": 3277
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.464128912630645e-05,
+      "loss": 3.1954,
+      "step": 3278
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.47837049338956e-05,
+      "loss": 4.3957,
+      "step": 3279
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.492615234210473e-05,
+      "loss": 3.6936,
+      "step": 3280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.506863105510476e-05,
+      "loss": 5.1745,
+      "step": 3281
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.521114077700108e-05,
+      "loss": 3.3525,
+      "step": 3282
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.535368121183502e-05,
+      "loss": 5.2428,
+      "step": 3283
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.549625206358409e-05,
+      "loss": 2.6518,
+      "step": 3284
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.563885303616255e-05,
+      "loss": 2.5005,
+      "step": 3285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.578148383342208e-05,
+      "loss": 4.3033,
+      "step": 3286
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.592414415915269e-05,
+      "loss": 4.5817,
+      "step": 3287
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.606683371708276e-05,
+      "loss": 4.0487,
+      "step": 3288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.620955221088005e-05,
+      "loss": 3.9343,
+      "step": 3289
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.635229934415228e-05,
+      "loss": 4.3601,
+      "step": 3290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.649507482044763e-05,
+      "loss": 4.1236,
+      "step": 3291
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.663787834325547e-05,
+      "loss": 4.6349,
+      "step": 3292
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.678070961600691e-05,
+      "loss": 5.4595,
+      "step": 3293
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.69235683420754e-05,
+      "loss": 4.3168,
+      "step": 3294
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.70664542247773e-05,
+      "loss": 5.1268,
+      "step": 3295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.720936696737287e-05,
+      "loss": 5.7067,
+      "step": 3296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.735230627306637e-05,
+      "loss": 3.9717,
+      "step": 3297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.749527184500674e-05,
+      "loss": 3.8964,
+      "step": 3298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.763826338628865e-05,
+      "loss": 5.0691,
+      "step": 3299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.778128059995272e-05,
+      "loss": 3.6417,
+      "step": 3300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.792432318898627e-05,
+      "loss": 4.7254,
+      "step": 3301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.806739085632382e-05,
+      "loss": 6.7784,
+      "step": 3302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.821048330484808e-05,
+      "loss": 3.6644,
+      "step": 3303
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.835360023739005e-05,
+      "loss": 3.9632,
+      "step": 3304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.849674135672994e-05,
+      "loss": 3.6139,
+      "step": 3305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.863990636559776e-05,
+      "loss": 5.3912,
+      "step": 3306
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.878309496667387e-05,
+      "loss": 4.7324,
+      "step": 3307
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.892630686258967e-05,
+      "loss": 3.8544,
+      "step": 3308
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.906954175592816e-05,
+      "loss": 3.8277,
+      "step": 3309
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.92127993492246e-05,
+      "loss": 4.0488,
+      "step": 3310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.935607934496699e-05,
+      "loss": 3.9911,
+      "step": 3311
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.949938144559709e-05,
+      "loss": 3.9882,
+      "step": 3312
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.96427053535106e-05,
+      "loss": 4.1758,
+      "step": 3313
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.97860507710577e-05,
+      "loss": 4.4127,
+      "step": 3314
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.992941740054421e-05,
+      "loss": 5.0134,
+      "step": 3315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.007280494423175e-05,
+      "loss": 4.4034,
+      "step": 3316
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.021621310433864e-05,
+      "loss": 5.5089,
+      "step": 3317
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.03596415830401e-05,
+      "loss": 4.0167,
+      "step": 3318
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.05030900824696e-05,
+      "loss": 4.051,
+      "step": 3319
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.064655830471866e-05,
+      "loss": 2.908,
+      "step": 3320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.079004595183796e-05,
+      "loss": 4.6911,
+      "step": 3321
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.093355272583784e-05,
+      "loss": 5.1059,
+      "step": 3322
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.107707832868894e-05,
+      "loss": 3.4827,
+      "step": 3323
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.122062246232273e-05,
+      "loss": 5.9564,
+      "step": 3324
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.136418482863226e-05,
+      "loss": 4.808,
+      "step": 3325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.15077651294727e-05,
+      "loss": 3.3827,
+      "step": 3326
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.165136306666188e-05,
+      "loss": 2.7355,
+      "step": 3327
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.179497834198125e-05,
+      "loss": 5.8803,
+      "step": 3328
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.193861065717611e-05,
+      "loss": 5.4122,
+      "step": 3329
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208225971395618e-05,
+      "loss": 5.0559,
+      "step": 3330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.222592521399669e-05,
+      "loss": 5.1709,
+      "step": 3331
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.236960685893861e-05,
+      "loss": 4.1244,
+      "step": 3332
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.251330435038936e-05,
+      "loss": 5.1592,
+      "step": 3333
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.265701738992344e-05,
+      "loss": 4.9642,
+      "step": 3334
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.280074567908325e-05,
+      "loss": 2.2403,
+      "step": 3335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.29444889193793e-05,
+      "loss": 4.2164,
+      "step": 3336
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.308824681229112e-05,
+      "loss": 4.8038,
+      "step": 3337
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.32320190592678e-05,
+      "loss": 5.9378,
+      "step": 3338
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.337580536172865e-05,
+      "loss": 5.6774,
+      "step": 3339
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.351960542106374e-05,
+      "loss": 6.0328,
+      "step": 3340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.366341893863464e-05,
+      "loss": 5.0482,
+      "step": 3341
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.38072456157749e-05,
+      "loss": 5.5857,
+      "step": 3342
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.395108515379073e-05,
+      "loss": 4.8085,
+      "step": 3343
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.40949372539618e-05,
+      "loss": 5.0001,
+      "step": 3344
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.423880161754161e-05,
+      "loss": 5.4827,
+      "step": 3345
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.438267794575794e-05,
+      "loss": 4.433,
+      "step": 3346
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.452656593981402e-05,
+      "loss": 4.0117,
+      "step": 3347
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.467046530088879e-05,
+      "loss": 3.4652,
+      "step": 3348
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.481437573013753e-05,
+      "loss": 5.4959,
+      "step": 3349
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.495829692869248e-05,
+      "loss": 4.3361,
+      "step": 3350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.510222859766377e-05,
+      "loss": 3.541,
+      "step": 3351
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.524617043813954e-05,
+      "loss": 4.9647,
+      "step": 3352
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.539012215118685e-05,
+      "loss": 4.8253,
+      "step": 3353
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.55340834378523e-05,
+      "loss": 4.8504,
+      "step": 3354
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.567805399916259e-05,
+      "loss": 3.8186,
+      "step": 3355
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.582203353612516e-05,
+      "loss": 3.7267,
+      "step": 3356
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.59660217497288e-05,
+      "loss": 5.0673,
+      "step": 3357
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.611001834094423e-05,
+      "loss": 5.0651,
+      "step": 3358
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.625402301072498e-05,
+      "loss": 5.8064,
+      "step": 3359
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.639803546000758e-05,
+      "loss": 3.4154,
+      "step": 3360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.654205538971255e-05,
+      "loss": 6.1819,
+      "step": 3361
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.668608250074466e-05,
+      "loss": 3.8677,
+      "step": 3362
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.683011649399397e-05,
+      "loss": 4.3792,
+      "step": 3363
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.697415707033621e-05,
+      "loss": 3.853,
+      "step": 3364
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.711820393063331e-05,
+      "loss": 5.5768,
+      "step": 3365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.726225677573444e-05,
+      "loss": 4.1216,
+      "step": 3366
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.740631530647607e-05,
+      "loss": 4.8083,
+      "step": 3367
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.755037922368295e-05,
+      "loss": 2.4917,
+      "step": 3368
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.769444822816863e-05,
+      "loss": 3.5717,
+      "step": 3369
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.783852202073611e-05,
+      "loss": 5.9384,
+      "step": 3370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.798260030217846e-05,
+      "loss": 5.3521,
+      "step": 3371
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.81266827732794e-05,
+      "loss": 4.3538,
+      "step": 3372
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.827076913481396e-05,
+      "loss": 4.9267,
+      "step": 3373
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.8414859087549e-05,
+      "loss": 5.2245,
+      "step": 3374
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.855895233224422e-05,
+      "loss": 4.336,
+      "step": 3375
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.870304856965213e-05,
+      "loss": 5.2084,
+      "step": 3376
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.884714750051931e-05,
+      "loss": 5.5713,
+      "step": 3377
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.899124882558641e-05,
+      "loss": 5.3865,
+      "step": 3378
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.913535224558939e-05,
+      "loss": 4.6344,
+      "step": 3379
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.92794574612598e-05,
+      "loss": 3.8206,
+      "step": 3380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.942356417332531e-05,
+      "loss": 4.7867,
+      "step": 3381
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.956767208251081e-05,
+      "loss": 4.6582,
+      "step": 3382
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.97117808895384e-05,
+      "loss": 4.4913,
+      "step": 3383
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.985589029512843e-05,
+      "loss": 3.343,
+      "step": 3384
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.999999999999998e-05,
+      "loss": 5.0676,
+      "step": 3385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010014410970487154,
+      "loss": 4.6097,
+      "step": 3386
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010028821911046159,
+      "loss": 3.7804,
+      "step": 3387
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010043232791748916,
+      "loss": 5.8385,
+      "step": 3388
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010057643582667467,
+      "loss": 4.5762,
+      "step": 3389
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010072054253874017,
+      "loss": 4.6987,
+      "step": 3390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010086464775441058,
+      "loss": 5.5676,
+      "step": 3391
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010100875117441356,
+      "loss": 4.1878,
+      "step": 3392
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010115285249948066,
+      "loss": 4.474,
+      "step": 3393
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010129695143034784,
+      "loss": 3.5306,
+      "step": 3394
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010144104766775575,
+      "loss": 3.1025,
+      "step": 3395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010158514091245098,
+      "loss": 5.1589,
+      "step": 3396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010172923086518602,
+      "loss": 4.548,
+      "step": 3397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010187331722672058,
+      "loss": 4.7529,
+      "step": 3398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010201739969782151,
+      "loss": 3.5474,
+      "step": 3399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010216147797926385,
+      "loss": 4.5425,
+      "step": 3400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010230555177183134,
+      "loss": 4.3315,
+      "step": 3401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010244962077631704,
+      "loss": 4.4152,
+      "step": 3402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010259368469352391,
+      "loss": 3.923,
+      "step": 3403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010273774322426553,
+      "loss": 6.0836,
+      "step": 3404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010288179606936666,
+      "loss": 4.8466,
+      "step": 3405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010302584292966377,
+      "loss": 4.9209,
+      "step": 3406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000103169883506006,
+      "loss": 4.8205,
+      "step": 3407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010331391749925531,
+      "loss": 4.6189,
+      "step": 3408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010345794461028742,
+      "loss": 5.0399,
+      "step": 3409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001036019645399924,
+      "loss": 4.2698,
+      "step": 3410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000103745976989275,
+      "loss": 6.1855,
+      "step": 3411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010388998165905574,
+      "loss": 5.9591,
+      "step": 3412
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010403397825027116,
+      "loss": 5.0729,
+      "step": 3413
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001041779664638748,
+      "loss": 5.2212,
+      "step": 3414
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010432194600083738,
+      "loss": 4.4422,
+      "step": 3415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010446591656214765,
+      "loss": 5.2095,
+      "step": 3416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001046098778488131,
+      "loss": 4.1132,
+      "step": 3417
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010475382956186044,
+      "loss": 5.5735,
+      "step": 3418
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010489777140233618,
+      "loss": 5.0976,
+      "step": 3419
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001050417030713075,
+      "loss": 3.1105,
+      "step": 3420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010518562426986245,
+      "loss": 3.0226,
+      "step": 3421
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010532953469911118,
+      "loss": 4.8451,
+      "step": 3422
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010547343406018594,
+      "loss": 5.0893,
+      "step": 3423
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010561732205424204,
+      "loss": 4.8718,
+      "step": 3424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010576119838245838,
+      "loss": 3.3639,
+      "step": 3425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010590506274603818,
+      "loss": 4.692,
+      "step": 3426
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010604891484620924,
+      "loss": 4.8307,
+      "step": 3427
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010619275438422507,
+      "loss": 3.87,
+      "step": 3428
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010633658106136533,
+      "loss": 3.257,
+      "step": 3429
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010648039457893624,
+      "loss": 4.8478,
+      "step": 3430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010662419463827134,
+      "loss": 3.9348,
+      "step": 3431
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010676798094073217,
+      "loss": 4.6664,
+      "step": 3432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010691175318770885,
+      "loss": 3.6669,
+      "step": 3433
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010705551108062067,
+      "loss": 4.421,
+      "step": 3434
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001071992543209167,
+      "loss": 4.0281,
+      "step": 3435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010734298261007653,
+      "loss": 4.3998,
+      "step": 3436
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010748669564961061,
+      "loss": 4.5756,
+      "step": 3437
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010763039314106138,
+      "loss": 3.1848,
+      "step": 3438
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010777407478600328,
+      "loss": 4.1359,
+      "step": 3439
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010791774028604377,
+      "loss": 5.5172,
+      "step": 3440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010806138934282386,
+      "loss": 3.1285,
+      "step": 3441
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001082050216580187,
+      "loss": 4.45,
+      "step": 3442
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001083486369333381,
+      "loss": 6.9383,
+      "step": 3443
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010849223487052726,
+      "loss": 4.924,
+      "step": 3444
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010863581517136771,
+      "loss": 4.4503,
+      "step": 3445
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010877937753767724,
+      "loss": 4.4152,
+      "step": 3446
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010892292167131103,
+      "loss": 5.0716,
+      "step": 3447
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010906644727416213,
+      "loss": 5.1099,
+      "step": 3448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010920995404816201,
+      "loss": 5.3624,
+      "step": 3449
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010935344169528132,
+      "loss": 4.863,
+      "step": 3450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010949690991753037,
+      "loss": 5.2895,
+      "step": 3451
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010964035841695985,
+      "loss": 4.4525,
+      "step": 3452
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010978378689566136,
+      "loss": 4.4526,
+      "step": 3453
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001099271950557682,
+      "loss": 3.1544,
+      "step": 3454
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011007058259945577,
+      "loss": 3.934,
+      "step": 3455
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011021394922894228,
+      "loss": 5.6657,
+      "step": 3456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011035729464648939,
+      "loss": 4.1867,
+      "step": 3457
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011050061855440288,
+      "loss": 5.61,
+      "step": 3458
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011064392065503297,
+      "loss": 5.1939,
+      "step": 3459
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011078720065077537,
+      "loss": 3.5784,
+      "step": 3460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001109304582440718,
+      "loss": 5.4499,
+      "step": 3461
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001110736931374103,
+      "loss": 3.4885,
+      "step": 3462
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011121690503332609,
+      "loss": 3.8599,
+      "step": 3463
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011136009363440221,
+      "loss": 4.3764,
+      "step": 3464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011150325864327003,
+      "loss": 4.9893,
+      "step": 3465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001116463997626099,
+      "loss": 3.9922,
+      "step": 3466
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001117895166951519,
+      "loss": 3.1652,
+      "step": 3467
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011193260914367616,
+      "loss": 5.5426,
+      "step": 3468
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001120756768110137,
+      "loss": 3.9636,
+      "step": 3469
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011221871940004724,
+      "loss": 4.9923,
+      "step": 3470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011236173661371132,
+      "loss": 4.4365,
+      "step": 3471
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011250472815499322,
+      "loss": 3.0209,
+      "step": 3472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001126476937269336,
+      "loss": 5.306,
+      "step": 3473
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011279063303262711,
+      "loss": 5.36,
+      "step": 3474
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011293354577522269,
+      "loss": 4.7377,
+      "step": 3475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011307643165792459,
+      "loss": 5.9739,
+      "step": 3476
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011321929038399306,
+      "loss": 5.5498,
+      "step": 3477
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001133621216567445,
+      "loss": 4.5507,
+      "step": 3478
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011350492517955233,
+      "loss": 3.4895,
+      "step": 3479
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011364770065584772,
+      "loss": 6.6102,
+      "step": 3480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011379044778911993,
+      "loss": 3.9488,
+      "step": 3481
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011393316628291724,
+      "loss": 3.4093,
+      "step": 3482
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011407585584084728,
+      "loss": 3.6574,
+      "step": 3483
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011421851616657789,
+      "loss": 3.1521,
+      "step": 3484
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011436114696383742,
+      "loss": 4.2922,
+      "step": 3485
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001145037479364159,
+      "loss": 4.0674,
+      "step": 3486
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011464631878816496,
+      "loss": 4.57,
+      "step": 3487
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011478885922299892,
+      "loss": 5.0056,
+      "step": 3488
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011493136894489523,
+      "loss": 4.8894,
+      "step": 3489
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011507384765789524,
+      "loss": 3.4355,
+      "step": 3490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011521629506610437,
+      "loss": 4.1676,
+      "step": 3491
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011535871087369352,
+      "loss": 3.3072,
+      "step": 3492
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001155010947848989,
+      "loss": 4.8217,
+      "step": 3493
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011564344650402308,
+      "loss": 5.6193,
+      "step": 3494
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001157857657354354,
+      "loss": 3.9896,
+      "step": 3495
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011592805218357275,
+      "loss": 4.1474,
+      "step": 3496
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011607030555294005,
+      "loss": 5.7826,
+      "step": 3497
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011621252554811095,
+      "loss": 3.7171,
+      "step": 3498
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011635471187372841,
+      "loss": 4.2329,
+      "step": 3499
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011649686423450521,
+      "loss": 3.6677,
+      "step": 3500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011663898233522493,
+      "loss": 3.9641,
+      "step": 3501
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011678106588074203,
+      "loss": 4.5444,
+      "step": 3502
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011692311457598286,
+      "loss": 4.3701,
+      "step": 3503
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001170651281259461,
+      "loss": 5.063,
+      "step": 3504
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011720710623570345,
+      "loss": 4.3469,
+      "step": 3505
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011734904861040026,
+      "loss": 4.4282,
+      "step": 3506
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011749095495525581,
+      "loss": 4.7957,
+      "step": 3507
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011763282497556464,
+      "loss": 4.0827,
+      "step": 3508
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011777465837669648,
+      "loss": 4.2944,
+      "step": 3509
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001179164548640971,
+      "loss": 5.6784,
+      "step": 3510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011805821414328897,
+      "loss": 4.8205,
+      "step": 3511
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011819993591987189,
+      "loss": 4.3436,
+      "step": 3512
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011834161989952343,
+      "loss": 4.1213,
+      "step": 3513
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011848326578799975,
+      "loss": 4.2297,
+      "step": 3514
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011862487329113607,
+      "loss": 5.3039,
+      "step": 3515
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011876644211484726,
+      "loss": 3.1185,
+      "step": 3516
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011890797196512878,
+      "loss": 2.9568,
+      "step": 3517
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001190494625480567,
+      "loss": 5.1184,
+      "step": 3518
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011919091356978889,
+      "loss": 5.2488,
+      "step": 3519
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001193323247365652,
+      "loss": 5.6665,
+      "step": 3520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011947369575470832,
+      "loss": 2.6448,
+      "step": 3521
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011961502633062445,
+      "loss": 4.9943,
+      "step": 3522
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011975631617080333,
+      "loss": 4.3696,
+      "step": 3523
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00011989756498181996,
+      "loss": 4.3503,
+      "step": 3524
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012003877247033405,
+      "loss": 4.5397,
+      "step": 3525
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012017993834309135,
+      "loss": 4.8294,
+      "step": 3526
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012032106230692393,
+      "loss": 4.6066,
+      "step": 3527
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012046214406875092,
+      "loss": 2.5631,
+      "step": 3528
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012060318333557914,
+      "loss": 4.9219,
+      "step": 3529
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012074417981450361,
+      "loss": 5.4107,
+      "step": 3530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012088513321270824,
+      "loss": 5.6068,
+      "step": 3531
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012102604323746632,
+      "loss": 3.8196,
+      "step": 3532
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012116690959614147,
+      "loss": 4.3714,
+      "step": 3533
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012130773199618773,
+      "loss": 4.8876,
+      "step": 3534
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012144851014515051,
+      "loss": 5.3628,
+      "step": 3535
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012158924375066712,
+      "loss": 4.0905,
+      "step": 3536
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012172993252046743,
+      "loss": 4.4735,
+      "step": 3537
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012187057616237443,
+      "loss": 4.5478,
+      "step": 3538
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012201117438430454,
+      "loss": 4.7086,
+      "step": 3539
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012215172689426898,
+      "loss": 3.4682,
+      "step": 3540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012229223340037363,
+      "loss": 4.8602,
+      "step": 3541
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001224326936108199,
+      "loss": 3.2764,
+      "step": 3542
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001225731072339054,
+      "loss": 4.9465,
+      "step": 3543
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012271347397802452,
+      "loss": 4.3252,
+      "step": 3544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012285379355166893,
+      "loss": 5.1438,
+      "step": 3545
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001229940656634283,
+      "loss": 4.5622,
+      "step": 3546
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001231342900219909,
+      "loss": 6.7697,
+      "step": 3547
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012327446633614406,
+      "loss": 4.6452,
+      "step": 3548
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001234145943147751,
+      "loss": 4.22,
+      "step": 3549
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012355467366687153,
+      "loss": 4.1246,
+      "step": 3550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012369470410152192,
+      "loss": 3.2321,
+      "step": 3551
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012383468532791635,
+      "loss": 5.4892,
+      "step": 3552
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012397461705534722,
+      "loss": 3.9886,
+      "step": 3553
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012411449899320975,
+      "loss": 3.5535,
+      "step": 3554
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012425433085100221,
+      "loss": 3.7592,
+      "step": 3555
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001243941123383274,
+      "loss": 3.0023,
+      "step": 3556
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012453384316489237,
+      "loss": 4.7482,
+      "step": 3557
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012467352304050948,
+      "loss": 4.5148,
+      "step": 3558
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012481315167509692,
+      "loss": 4.1078,
+      "step": 3559
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012495272877867926,
+      "loss": 4.5267,
+      "step": 3560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001250922540613881,
+      "loss": 3.5773,
+      "step": 3561
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001252317272334627,
+      "loss": 3.5862,
+      "step": 3562
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001253711480052504,
+      "loss": 4.3628,
+      "step": 3563
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012551051608720762,
+      "loss": 4.8142,
+      "step": 3564
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012564983118990003,
+      "loss": 4.0778,
+      "step": 3565
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012578909302400328,
+      "loss": 3.9881,
+      "step": 3566
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012592830130030375,
+      "loss": 4.5852,
+      "step": 3567
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000126067455729699,
+      "loss": 6.0347,
+      "step": 3568
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012620655602319848,
+      "loss": 3.2074,
+      "step": 3569
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012634560189192393,
+      "loss": 3.8046,
+      "step": 3570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012648459304711026,
+      "loss": 3.5718,
+      "step": 3571
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001266235292001059,
+      "loss": 5.7258,
+      "step": 3572
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012676241006237361,
+      "loss": 6.1569,
+      "step": 3573
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012690123534549094,
+      "loss": 4.992,
+      "step": 3574
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001270400047611508,
+      "loss": 4.2766,
+      "step": 3575
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012717871802116217,
+      "loss": 4.6844,
+      "step": 3576
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012731737483745068,
+      "loss": 5.2129,
+      "step": 3577
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00012745597492205913,
+      "loss": 4.3511,
+      "step": 3578
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001275945179871481,
+      "loss": 4.5108,
+      "step": 3579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001277330037449968,
+      "loss": 4.6735,
+      "step": 3580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012787143190800323,
+      "loss": 4.7558,
+      "step": 3581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012800980218868503,
+      "loss": 4.7622,
+      "step": 3582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012814811429968017,
+      "loss": 4.4739,
+      "step": 3583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012828636795374728,
+      "loss": 4.9888,
+      "step": 3584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012842456286376645,
+      "loss": 4.7299,
+      "step": 3585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001285626987427398,
+      "loss": 5.2165,
+      "step": 3586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012870077530379202,
+      "loss": 5.5194,
+      "step": 3587
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012883879226017098,
+      "loss": 4.117,
+      "step": 3588
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012897674932524833,
+      "loss": 2.9455,
+      "step": 3589
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012911464621252007,
+      "loss": 3.5469,
+      "step": 3590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012925248263560732,
+      "loss": 5.453,
+      "step": 3591
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012939025830825657,
+      "loss": 4.2363,
+      "step": 3592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001295279729443406,
+      "loss": 4.5015,
+      "step": 3593
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001296656262578589,
+      "loss": 4.7498,
+      "step": 3594
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001298032179629383,
+      "loss": 4.949,
+      "step": 3595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00012994074777383373,
+      "loss": 4.0033,
+      "step": 3596
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013007821540492846,
+      "loss": 4.4827,
+      "step": 3597
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013021562057073495,
+      "loss": 4.5794,
+      "step": 3598
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013035296298589543,
+      "loss": 3.9002,
+      "step": 3599
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001304902423651824,
+      "loss": 3.1388,
+      "step": 3600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001306274584234993,
+      "loss": 4.8357,
+      "step": 3601
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013076461087588109,
+      "loss": 4.6211,
+      "step": 3602
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013090169943749474,
+      "loss": 4.4545,
+      "step": 3603
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013103872382364,
+      "loss": 6.2125,
+      "step": 3604
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013117568374974986,
+      "loss": 5.9519,
+      "step": 3605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013131257893139118,
+      "loss": 3.4335,
+      "step": 3606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013144940908426528,
+      "loss": 4.5532,
+      "step": 3607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013158617392420856,
+      "loss": 4.7403,
+      "step": 3608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013172287316719295,
+      "loss": 3.1718,
+      "step": 3609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013185950652932674,
+      "loss": 4.3693,
+      "step": 3610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001319960737268549,
+      "loss": 5.7052,
+      "step": 3611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013213257447616006,
+      "loss": 2.8964,
+      "step": 3612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013226900849376257,
+      "loss": 4.2768,
+      "step": 3613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013240537549632147,
+      "loss": 5.4271,
+      "step": 3614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013254167520063506,
+      "loss": 4.38,
+      "step": 3615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013267790732364126,
+      "loss": 2.9476,
+      "step": 3616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013281407158241838,
+      "loss": 4.3242,
+      "step": 3617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013295016769418575,
+      "loss": 4.0902,
+      "step": 3618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013308619537630414,
+      "loss": 4.1201,
+      "step": 3619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013322215434627646,
+      "loss": 4.8524,
+      "step": 3620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001333580443217483,
+      "loss": 5.176,
+      "step": 3621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013349386502050856,
+      "loss": 4.3915,
+      "step": 3622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013362961616049003,
+      "loss": 6.4784,
+      "step": 3623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013376529745976994,
+      "loss": 4.2919,
+      "step": 3624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013390090863657047,
+      "loss": 4.3072,
+      "step": 3625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001340364494092595,
+      "loss": 3.3838,
+      "step": 3626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013417191949635133,
+      "loss": 4.93,
+      "step": 3627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001343073186165067,
+      "loss": 5.3159,
+      "step": 3628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001344426464885339,
+      "loss": 5.0429,
+      "step": 3629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013457790283138922,
+      "loss": 5.7544,
+      "step": 3630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013471308736417745,
+      "loss": 2.9326,
+      "step": 3631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001348481998061525,
+      "loss": 4.8496,
+      "step": 3632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000134983239876718,
+      "loss": 4.9135,
+      "step": 3633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001351182072954279,
+      "loss": 5.6147,
+      "step": 3634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013525310178198705,
+      "loss": 3.5088,
+      "step": 3635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013538792305625174,
+      "loss": 3.7763,
+      "step": 3636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013552267083823027,
+      "loss": 4.462,
+      "step": 3637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013565734484808362,
+      "loss": 4.8953,
+      "step": 3638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001357919448061259,
+      "loss": 4.6085,
+      "step": 3639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013592647043282513,
+      "loss": 5.9194,
+      "step": 3640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001360609214488036,
+      "loss": 6.4714,
+      "step": 3641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013619529757483847,
+      "loss": 5.8612,
+      "step": 3642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013632959853186275,
+      "loss": 5.0752,
+      "step": 3643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013646382404096526,
+      "loss": 4.7088,
+      "step": 3644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013659797382339158,
+      "loss": 5.0873,
+      "step": 3645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013673204760054452,
+      "loss": 3.3122,
+      "step": 3646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013686604509398483,
+      "loss": 4.2588,
+      "step": 3647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001369999660254317,
+      "loss": 5.1305,
+      "step": 3648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013713381011676312,
+      "loss": 3.4471,
+      "step": 3649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013726757709001689,
+      "loss": 2.9806,
+      "step": 3650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013740126666739083,
+      "loss": 5.816,
+      "step": 3651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001375348785712436,
+      "loss": 4.9762,
+      "step": 3652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000137668412524095,
+      "loss": 3.5537,
+      "step": 3653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013780186824862694,
+      "loss": 3.6946,
+      "step": 3654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013793524546768356,
+      "loss": 6.1408,
+      "step": 3655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013806854390427224,
+      "loss": 4.998,
+      "step": 3656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013820176328156384,
+      "loss": 4.255,
+      "step": 3657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013833490332289334,
+      "loss": 3.6327,
+      "step": 3658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001384679637517608,
+      "loss": 3.631,
+      "step": 3659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013860094429183131,
+      "loss": 5.5136,
+      "step": 3660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013873384466693597,
+      "loss": 3.8599,
+      "step": 3661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001388666646010723,
+      "loss": 4.3145,
+      "step": 3662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001388666646010723,
+      "loss": 3.7984,
+      "step": 3663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013899940381840507,
+      "loss": 5.4326,
+      "step": 3664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013913206204326642,
+      "loss": 4.2381,
+      "step": 3665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013926463900015688,
+      "loss": 5.4115,
+      "step": 3666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013939713441374566,
+      "loss": 1.5954,
+      "step": 3667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013952954800887138,
+      "loss": 4.4147,
+      "step": 3668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00013966187951054254,
+      "loss": 5.3612,
+      "step": 3669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001397941286439381,
+      "loss": 5.5396,
+      "step": 3670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001399262951344082,
+      "loss": 3.944,
+      "step": 3671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001400583787074744,
+      "loss": 4.907,
+      "step": 3672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001401903790888307,
+      "loss": 3.7602,
+      "step": 3673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014032229600434367,
+      "loss": 4.7195,
+      "step": 3674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014045412918005324,
+      "loss": 5.0334,
+      "step": 3675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014058587834217353,
+      "loss": 3.809,
+      "step": 3676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014071754321709276,
+      "loss": 4.0565,
+      "step": 3677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014084912353137439,
+      "loss": 4.9664,
+      "step": 3678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014098061901175743,
+      "loss": 5.7509,
+      "step": 3679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001411120293851571,
+      "loss": 3.6921,
+      "step": 3680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014124335437866537,
+      "loss": 4.2216,
+      "step": 3681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001413745937195515,
+      "loss": 2.8169,
+      "step": 3682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014150574713526264,
+      "loss": 3.7806,
+      "step": 3683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001416368143534244,
+      "loss": 3.745,
+      "step": 3684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014176779510184138,
+      "loss": 5.2212,
+      "step": 3685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014189868910849778,
+      "loss": 5.0099,
+      "step": 3686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001420294961015579,
+      "loss": 5.9999,
+      "step": 3687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001421602158093668,
+      "loss": 6.044,
+      "step": 3688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001422908479604508,
+      "loss": 4.7114,
+      "step": 3689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014242139228351803,
+      "loss": 5.7112,
+      "step": 3690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014255184850745894,
+      "loss": 3.6468,
+      "step": 3691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014268221636134717,
+      "loss": 4.5269,
+      "step": 3692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014281249557443972,
+      "loss": 4.2088,
+      "step": 3693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014294268587617765,
+      "loss": 4.3622,
+      "step": 3694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014307278699618673,
+      "loss": 4.4058,
+      "step": 3695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014320279866427796,
+      "loss": 4.3452,
+      "step": 3696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000143332720610448,
+      "loss": 4.4713,
+      "step": 3697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014346255256488,
+      "loss": 2.979,
+      "step": 3698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014359229425794386,
+      "loss": 5.2926,
+      "step": 3699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014372194542019705,
+      "loss": 3.8619,
+      "step": 3700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001438515057823849,
+      "loss": 3.3441,
+      "step": 3701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001439809750754415,
+      "loss": 5.0875,
+      "step": 3702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001441103530304899,
+      "loss": 4.6533,
+      "step": 3703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014423963937884295,
+      "loss": 3.7304,
+      "step": 3704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001443688338520037,
+      "loss": 4.208,
+      "step": 3705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014449793618166588,
+      "loss": 4.3856,
+      "step": 3706
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001446269460997149,
+      "loss": 3.6892,
+      "step": 3707
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001447558633382278,
+      "loss": 3.9713,
+      "step": 3708
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001448846876294742,
+      "loss": 5.4441,
+      "step": 3709
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014501341870591677,
+      "loss": 4.3776,
+      "step": 3710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014514205630021166,
+      "loss": 4.2295,
+      "step": 3711
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014527060014520927,
+      "loss": 5.9217,
+      "step": 3712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014539904997395468,
+      "loss": 4.6488,
+      "step": 3713
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014552740551968812,
+      "loss": 5.3394,
+      "step": 3714
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014565566651584576,
+      "loss": 5.5488,
+      "step": 3715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014578383269606005,
+      "loss": 4.7573,
+      "step": 3716
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014591190379416034,
+      "loss": 3.5608,
+      "step": 3717
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014603987954417348,
+      "loss": 3.2849,
+      "step": 3718
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014616775968032437,
+      "loss": 6.344,
+      "step": 3719
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014629554393703637,
+      "loss": 3.1915,
+      "step": 3720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014642323204893203,
+      "loss": 4.7288,
+      "step": 3721
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014655082375083353,
+      "loss": 3.8621,
+      "step": 3722
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001466783187777634,
+      "loss": 2.3768,
+      "step": 3723
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014680571686494478,
+      "loss": 3.704,
+      "step": 3724
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001469330177478022,
+      "loss": 6.7695,
+      "step": 3725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014706022116196203,
+      "loss": 3.1252,
+      "step": 3726
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001471873268432531,
+      "loss": 4.3141,
+      "step": 3727
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001473143345277072,
+      "loss": 4.2864,
+      "step": 3728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001474412439515596,
+      "loss": 5.016,
+      "step": 3729
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014756805485124968,
+      "loss": 4.3795,
+      "step": 3730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014769476696342137,
+      "loss": 3.9624,
+      "step": 3731
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014782138002492385,
+      "loss": 3.7814,
+      "step": 3732
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014794789377281192,
+      "loss": 2.8949,
+      "step": 3733
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001480743079443467,
+      "loss": 5.1556,
+      "step": 3734
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014820062227699602,
+      "loss": 4.7006,
+      "step": 3735
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014832683650843511,
+      "loss": 5.5222,
+      "step": 3736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001484529503765471,
+      "loss": 3.3014,
+      "step": 3737
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014857896361942342,
+      "loss": 3.2257,
+      "step": 3738
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014870487597536483,
+      "loss": 4.572,
+      "step": 3739
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014883068718288115,
+      "loss": 5.0811,
+      "step": 3740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014895639698069257,
+      "loss": 3.8951,
+      "step": 3741
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014908200510772973,
+      "loss": 3.3917,
+      "step": 3742
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001492075113031345,
+      "loss": 3.9417,
+      "step": 3743
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014933291530626043,
+      "loss": 4.0961,
+      "step": 3744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014945821685667326,
+      "loss": 3.2918,
+      "step": 3745
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014958341569415147,
+      "loss": 5.5683,
+      "step": 3746
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014970851155868698,
+      "loss": 5.4924,
+      "step": 3747
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00014983350419048544,
+      "loss": 2.8459,
+      "step": 3748
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001499583933299669,
+      "loss": 4.1097,
+      "step": 3749
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015008317871776644,
+      "loss": 3.7843,
+      "step": 3750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015020786009473444,
+      "loss": 4.7556,
+      "step": 3751
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001503324372019375,
+      "loss": 3.3994,
+      "step": 3752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015045690978065853,
+      "loss": 6.1826,
+      "step": 3753
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015058127757239763,
+      "loss": 4.5348,
+      "step": 3754
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015070554031887268,
+      "loss": 3.6316,
+      "step": 3755
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015082969776201942,
+      "loss": 3.9401,
+      "step": 3756
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015095374964399247,
+      "loss": 5.083,
+      "step": 3757
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001510776957071656,
+      "loss": 4.5172,
+      "step": 3758
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001512015356941324,
+      "loss": 3.7414,
+      "step": 3759
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001513252693477067,
+      "loss": 4.8879,
+      "step": 3760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015144889641092313,
+      "loss": 5.4126,
+      "step": 3761
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00015157241662703768,
+      "loss": 4.6521,
+      "step": 3762
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015169582973952853,
+      "loss": 4.6418,
+      "step": 3763
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015181913549209582,
+      "loss": 4.419,
+      "step": 3764
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015194233362866293,
+      "loss": 5.2597,
+      "step": 3765
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001520654238933767,
+      "loss": 3.6087,
+      "step": 3766
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015218840603060797,
+      "loss": 4.8273,
+      "step": 3767
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001523112797849521,
+      "loss": 5.2461,
+      "step": 3768
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015243404490122956,
+      "loss": 5.4128,
+      "step": 3769
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015255670112448652,
+      "loss": 4.1337,
+      "step": 3770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015267924819999524,
+      "loss": 4.2706,
+      "step": 3771
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015280168587325456,
+      "loss": 5.2054,
+      "step": 3772
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015292401388999064,
+      "loss": 4.3048,
+      "step": 3773
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015304623199615734,
+      "loss": 5.0687,
+      "step": 3774
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015316833993793674,
+      "loss": 2.8378,
+      "step": 3775
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015329033746173972,
+      "loss": 5.1788,
+      "step": 3776
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015341222431420652,
+      "loss": 3.4636,
+      "step": 3777
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015353400024220712,
+      "loss": 3.6784,
+      "step": 3778
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001536556649928421,
+      "loss": 4.7616,
+      "step": 3779
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015377721831344256,
+      "loss": 3.4698,
+      "step": 3780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015389865995157128,
+      "loss": 4.1419,
+      "step": 3781
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015401998965502292,
+      "loss": 3.8599,
+      "step": 3782
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015414120717182456,
+      "loss": 5.3911,
+      "step": 3783
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015426231225023634,
+      "loss": 3.798,
+      "step": 3784
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001543833046387517,
+      "loss": 5.4339,
+      "step": 3785
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015450418408609855,
+      "loss": 4.5079,
+      "step": 3786
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001546249503412389,
+      "loss": 2.2191,
+      "step": 3787
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015474560315337002,
+      "loss": 4.4164,
+      "step": 3788
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001548661422719248,
+      "loss": 3.3189,
+      "step": 3789
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015498656744657215,
+      "loss": 4.5827,
+      "step": 3790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015510687842721773,
+      "loss": 5.0949,
+      "step": 3791
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015522707496400427,
+      "loss": 4.8823,
+      "step": 3792
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015534715680731216,
+      "loss": 4.8666,
+      "step": 3793
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015546712370775997,
+      "loss": 4.0386,
+      "step": 3794
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015558697541620528,
+      "loss": 4.298,
+      "step": 3795
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015570671168374438,
+      "loss": 5.2583,
+      "step": 3796
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015582633226171366,
+      "loss": 4.1371,
+      "step": 3797
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015594583690168966,
+      "loss": 5.0338,
+      "step": 3798
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001560652253554897,
+      "loss": 4.3965,
+      "step": 3799
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015618449737517245,
+      "loss": 5.827,
+      "step": 3800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001563036527130382,
+      "loss": 4.0977,
+      "step": 3801
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001564226911216299,
+      "loss": 3.907,
+      "step": 3802
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015654161235373305,
+      "loss": 5.0189,
+      "step": 3803
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015666041616237657,
+      "loss": 4.8459,
+      "step": 3804
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015677910230083323,
+      "loss": 4.4777,
+      "step": 3805
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001568976705226202,
+      "loss": 4.2369,
+      "step": 3806
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015701612058149957,
+      "loss": 3.3705,
+      "step": 3807
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015713445223147874,
+      "loss": 4.4958,
+      "step": 3808
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001572526652268111,
+      "loss": 4.0231,
+      "step": 3809
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015737075932199628,
+      "loss": 6.5567,
+      "step": 3810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015748873427178127,
+      "loss": 4.2721,
+      "step": 3811
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015760658983115996,
+      "loss": 3.8553,
+      "step": 3812
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015772432575537448,
+      "loss": 4.8912,
+      "step": 3813
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015784194179991535,
+      "loss": 4.3134,
+      "step": 3814
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015795943772052212,
+      "loss": 3.9167,
+      "step": 3815
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015807681327318374,
+      "loss": 4.4445,
+      "step": 3816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015819406821413902,
+      "loss": 4.5259,
+      "step": 3817
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015831120229987757,
+      "loss": 2.8495,
+      "step": 3818
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015842821528713974,
+      "loss": 3.7925,
+      "step": 3819
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001585451069329174,
+      "loss": 3.8808,
+      "step": 3820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015866187699445448,
+      "loss": 5.5974,
+      "step": 3821
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001587785252292473,
+      "loss": 4.7161,
+      "step": 3822
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015889505139504535,
+      "loss": 5.6482,
+      "step": 3823
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015901145524985151,
+      "loss": 5.1727,
+      "step": 3824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001591277365519227,
+      "loss": 3.7927,
+      "step": 3825
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001592438950597703,
+      "loss": 3.9716,
+      "step": 3826
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015935993053216093,
+      "loss": 3.4954,
+      "step": 3827
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001594758427281164,
+      "loss": 5.0612,
+      "step": 3828
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015959163140691473,
+      "loss": 4.8297,
+      "step": 3829
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015970729632809044,
+      "loss": 4.0975,
+      "step": 3830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015982283725143501,
+      "loss": 4.4476,
+      "step": 3831
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00015993825393699742,
+      "loss": 3.9302,
+      "step": 3832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016005354614508484,
+      "loss": 5.7115,
+      "step": 3833
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016016871363626274,
+      "loss": 3.0797,
+      "step": 3834
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016028375617135567,
+      "loss": 4.9284,
+      "step": 3835
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016039867351144772,
+      "loss": 5.7349,
+      "step": 3836
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016051346541788296,
+      "loss": 3.4629,
+      "step": 3837
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001606281316522659,
+      "loss": 4.7699,
+      "step": 3838
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016074267197646212,
+      "loss": 4.2414,
+      "step": 3839
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001608570861525986,
+      "loss": 5.2755,
+      "step": 3840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016097137394306435,
+      "loss": 4.2651,
+      "step": 3841
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000161085535110511,
+      "loss": 3.8079,
+      "step": 3842
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016119956941785293,
+      "loss": 4.4519,
+      "step": 3843
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016131347662826792,
+      "loss": 3.7064,
+      "step": 3844
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016142725650519795,
+      "loss": 4.5315,
+      "step": 3845
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016154090881234933,
+      "loss": 3.5721,
+      "step": 3846
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016165443331369318,
+      "loss": 2.6599,
+      "step": 3847
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016176782977346622,
+      "loss": 3.8611,
+      "step": 3848
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001618810979561711,
+      "loss": 3.2625,
+      "step": 3849
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016199423762657667,
+      "loss": 6.2944,
+      "step": 3850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016210724854971885,
+      "loss": 3.9738,
+      "step": 3851
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016222013049090083,
+      "loss": 6.2986,
+      "step": 3852
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016233288321569375,
+      "loss": 3.9344,
+      "step": 3853
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016244550648993706,
+      "loss": 2.9729,
+      "step": 3854
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016255800007973902,
+      "loss": 3.2278,
+      "step": 3855
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016267036375147725,
+      "loss": 4.7092,
+      "step": 3856
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016278259727179912,
+      "loss": 4.7517,
+      "step": 3857
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016289470040762246,
+      "loss": 3.1298,
+      "step": 3858
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016300667292613573,
+      "loss": 4.3439,
+      "step": 3859
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016311851459479862,
+      "loss": 2.3499,
+      "step": 3860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016323022518134264,
+      "loss": 3.9978,
+      "step": 3861
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016334180445377156,
+      "loss": 5.7517,
+      "step": 3862
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016345325218036178,
+      "loss": 2.9722,
+      "step": 3863
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016356456812966283,
+      "loss": 3.7517,
+      "step": 3864
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016367575207049822,
+      "loss": 4.6417,
+      "step": 3865
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001637868037719652,
+      "loss": 3.7043,
+      "step": 3866
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016389772300343594,
+      "loss": 5.5638,
+      "step": 3867
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001640085095345576,
+      "loss": 4.6755,
+      "step": 3868
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001641191631352529,
+      "loss": 6.1308,
+      "step": 3869
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016422968357572072,
+      "loss": 4.8302,
+      "step": 3870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001643400706264364,
+      "loss": 2.9727,
+      "step": 3871
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001644503240581524,
+      "loss": 3.7745,
+      "step": 3872
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016456044364189843,
+      "loss": 6.3383,
+      "step": 3873
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016467042914898255,
+      "loss": 4.0456,
+      "step": 3874
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016478028035099107,
+      "loss": 4.7095,
+      "step": 3875
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016488999701978906,
+      "loss": 5.0763,
+      "step": 3876
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001649995789275212,
+      "loss": 2.7265,
+      "step": 3877
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016510902584661197,
+      "loss": 2.277,
+      "step": 3878
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016521833754976618,
+      "loss": 2.8167,
+      "step": 3879
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016532751380996942,
+      "loss": 3.7581,
+      "step": 3880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016543655440048874,
+      "loss": 3.5834,
+      "step": 3881
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001655454590948727,
+      "loss": 5.6007,
+      "step": 3882
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001656542276669523,
+      "loss": 2.5189,
+      "step": 3883
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000165762859890841,
+      "loss": 5.6656,
+      "step": 3884
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016587135554093559,
+      "loss": 4.4601,
+      "step": 3885
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016597971439191646,
+      "loss": 2.7559,
+      "step": 3886
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001660879362187481,
+      "loss": 3.9564,
+      "step": 3887
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016619602079667956,
+      "loss": 4.4333,
+      "step": 3888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016630396790124483,
+      "loss": 4.1894,
+      "step": 3889
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016641177730826362,
+      "loss": 2.965,
+      "step": 3890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001665194487938415,
+      "loss": 4.2556,
+      "step": 3891
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001666269821343703,
+      "loss": 5.252,
+      "step": 3892
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016673437710652893,
+      "loss": 4.8798,
+      "step": 3893
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016684163348728364,
+      "loss": 3.1314,
+      "step": 3894
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016694875105388848,
+      "loss": 4.6943,
+      "step": 3895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001670557295838857,
+      "loss": 3.3965,
+      "step": 3896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016716256885510652,
+      "loss": 5.3574,
+      "step": 3897
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001672692686456712,
+      "loss": 4.2825,
+      "step": 3898
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016737582873398957,
+      "loss": 4.1957,
+      "step": 3899
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016748224889876186,
+      "loss": 3.9085,
+      "step": 3900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016758852891897864,
+      "loss": 2.6038,
+      "step": 3901
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016769466857392167,
+      "loss": 4.4737,
+      "step": 3902
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016780066764316413,
+      "loss": 3.0688,
+      "step": 3903
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016790652590657122,
+      "loss": 5.6836,
+      "step": 3904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016801224314430058,
+      "loss": 3.3608,
+      "step": 3905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016811781913680265,
+      "loss": 2.9374,
+      "step": 3906
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016822325366482136,
+      "loss": 4.8676,
+      "step": 3907
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016832854650939414,
+      "loss": 4.0824,
+      "step": 3908
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001684336974518529,
+      "loss": 3.9287,
+      "step": 3909
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016853870627382418,
+      "loss": 4.6157,
+      "step": 3910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016864357275722962,
+      "loss": 4.4512,
+      "step": 3911
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016874829668428665,
+      "loss": 2.9588,
+      "step": 3912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001688528778375085,
+      "loss": 5.1217,
+      "step": 3913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016895731599970512,
+      "loss": 3.7758,
+      "step": 3914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016906161095398329,
+      "loss": 3.1388,
+      "step": 3915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016916576248374719,
+      "loss": 5.4236,
+      "step": 3916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016926977037269895,
+      "loss": 4.8247,
+      "step": 3917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016937363440483894,
+      "loss": 4.642,
+      "step": 3918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016947735436446634,
+      "loss": 4.5354,
+      "step": 3919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016958093003617938,
+      "loss": 2.9447,
+      "step": 3920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016968436120487623,
+      "loss": 3.1554,
+      "step": 3921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016978764765575495,
+      "loss": 6.0597,
+      "step": 3922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001698907891743142,
+      "loss": 5.5029,
+      "step": 3923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00016999378554635353,
+      "loss": 4.9624,
+      "step": 3924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017009663655797404,
+      "loss": 4.0446,
+      "step": 3925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001701993419955787,
+      "loss": 4.6816,
+      "step": 3926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017030190164587272,
+      "loss": 3.8372,
+      "step": 3927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017040431529586424,
+      "loss": 6.7613,
+      "step": 3928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001705065827328645,
+      "loss": 4.3798,
+      "step": 3929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017060870374448835,
+      "loss": 3.251,
+      "step": 3930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017071067811865476,
+      "loss": 3.7129,
+      "step": 3931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017081250564358725,
+      "loss": 4.1398,
+      "step": 3932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001709141861078144,
+      "loss": 3.8666,
+      "step": 3933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017101571930016997,
+      "loss": 4.9042,
+      "step": 3934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017111710500979378,
+      "loss": 4.8956,
+      "step": 3935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017121834302613183,
+      "loss": 4.0948,
+      "step": 3936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017131943313893695,
+      "loss": 5.7298,
+      "step": 3937
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017142037513826898,
+      "loss": 2.9018,
+      "step": 3938
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001715211688144955,
+      "loss": 3.09,
+      "step": 3939
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017162181395829205,
+      "loss": 3.748,
+      "step": 3940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001717223103606426,
+      "loss": 3.9221,
+      "step": 3941
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017182265781284,
+      "loss": 4.3505,
+      "step": 3942
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017192285610648656,
+      "loss": 4.1584,
+      "step": 3943
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017202290503349433,
+      "loss": 4.7341,
+      "step": 3944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017212280438608552,
+      "loss": 4.2031,
+      "step": 3945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00017222255395679293,
+      "loss": 3.6336,
+      "step": 3946
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017232215353846052,
+      "loss": 3.1931,
+      "step": 3947
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001724216029242436,
+      "loss": 3.6412,
+      "step": 3948
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017252090190760953,
+      "loss": 4.8932,
+      "step": 3949
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017262005028233804,
+      "loss": 4.3737,
+      "step": 3950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017271904784252145,
+      "loss": 6.3005,
+      "step": 3951
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017281789438256543,
+      "loss": 4.4431,
+      "step": 3952
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001729165896971894,
+      "loss": 4.1861,
+      "step": 3953
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017301513358142654,
+      "loss": 4.1152,
+      "step": 3954
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017311352583062478,
+      "loss": 5.3393,
+      "step": 3955
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017321176624044687,
+      "loss": 3.5945,
+      "step": 3956
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001733098546068708,
+      "loss": 4.8496,
+      "step": 3957
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017340779072619043,
+      "loss": 3.9284,
+      "step": 3958
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017350557439501572,
+      "loss": 4.5921,
+      "step": 3959
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001736032054102734,
+      "loss": 3.4852,
+      "step": 3960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017370068356920704,
+      "loss": 4.2946,
+      "step": 3961
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017379800866937773,
+      "loss": 4.8286,
+      "step": 3962
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001738951805086644,
+      "loss": 3.366,
+      "step": 3963
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017399219888526436,
+      "loss": 3.4029,
+      "step": 3964
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001740890635976935,
+      "loss": 4.1609,
+      "step": 3965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017418577444478688,
+      "loss": 4.7297,
+      "step": 3966
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017428233122569905,
+      "loss": 4.5518,
+      "step": 3967
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001743787337399047,
+      "loss": 4.4324,
+      "step": 3968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017447498178719878,
+      "loss": 3.8451,
+      "step": 3969
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001745710751676969,
+      "loss": 7.1251,
+      "step": 3970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017466701368183606,
+      "loss": 4.0376,
+      "step": 3971
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017476279713037494,
+      "loss": 3.8046,
+      "step": 3972
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017485842531439393,
+      "loss": 5.8115,
+      "step": 3973
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001749538980352961,
+      "loss": 3.8579,
+      "step": 3974
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017504921509480748,
+      "loss": 5.0765,
+      "step": 3975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001751443762949772,
+      "loss": 4.4722,
+      "step": 3976
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017523938143817807,
+      "loss": 3.1429,
+      "step": 3977
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017533423032710705,
+      "loss": 3.3321,
+      "step": 3978
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017542892276478563,
+      "loss": 3.1577,
+      "step": 3979
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017552345855456015,
+      "loss": 4.6531,
+      "step": 3980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001756178375001023,
+      "loss": 3.9065,
+      "step": 3981
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017571205940540948,
+      "loss": 3.7821,
+      "step": 3982
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017580612407480518,
+      "loss": 3.8436,
+      "step": 3983
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017590003131293962,
+      "loss": 3.0831,
+      "step": 3984
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017599378092478975,
+      "loss": 3.5197,
+      "step": 3985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017608737271566002,
+      "loss": 4.1434,
+      "step": 3986
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017618080649118252,
+      "loss": 5.005,
+      "step": 3987
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017627408205731765,
+      "loss": 4.1281,
+      "step": 3988
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001763671992203541,
+      "loss": 5.7449,
+      "step": 3989
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001764601577869098,
+      "loss": 4.8106,
+      "step": 3990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017655295756393197,
+      "loss": 4.1857,
+      "step": 3991
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017664559835869762,
+      "loss": 4.5627,
+      "step": 3992
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001767380799788138,
+      "loss": 4.3731,
+      "step": 3993
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017683040223221826,
+      "loss": 5.4277,
+      "step": 3994
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017692256492717967,
+      "loss": 3.6505,
+      "step": 3995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017701456787229804,
+      "loss": 3.9691,
+      "step": 3996
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017710641087650522,
+      "loss": 5.0168,
+      "step": 3997
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017719809374906512,
+      "loss": 2.7376,
+      "step": 3998
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017728961629957424,
+      "loss": 3.4168,
+      "step": 3999
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017738097833796212,
+      "loss": 3.6247,
+      "step": 4000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017747217967449154,
+      "loss": 3.2741,
+      "step": 4001
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000177563220119759,
+      "loss": 5.0797,
+      "step": 4002
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017765409948469527,
+      "loss": 5.4012,
+      "step": 4003
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017774481758056556,
+      "loss": 3.6824,
+      "step": 4004
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017783537421896988,
+      "loss": 4.4171,
+      "step": 4005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017792576921184366,
+      "loss": 3.7712,
+      "step": 4006
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001780160023714582,
+      "loss": 3.7085,
+      "step": 4007
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001781060735104206,
+      "loss": 3.7814,
+      "step": 4008
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001781959824416746,
+      "loss": 4.2938,
+      "step": 4009
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017828572897850072,
+      "loss": 3.9314,
+      "step": 4010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017837531293451685,
+      "loss": 3.6287,
+      "step": 4011
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017846473412367844,
+      "loss": 5.3297,
+      "step": 4012
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001785539923602789,
+      "loss": 4.4367,
+      "step": 4013
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017864308745895024,
+      "loss": 4.5247,
+      "step": 4014
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001787320192346631,
+      "loss": 4.6043,
+      "step": 4015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001788207875027274,
+      "loss": 3.0481,
+      "step": 4016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017890939207879257,
+      "loss": 4.6819,
+      "step": 4017
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017899783277884805,
+      "loss": 3.9603,
+      "step": 4018
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017908610941922356,
+      "loss": 4.7836,
+      "step": 4019
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017917422181658958,
+      "loss": 4.8263,
+      "step": 4020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017926216978795756,
+      "loss": 4.2645,
+      "step": 4021
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001793499531506805,
+      "loss": 5.36,
+      "step": 4022
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017943757172245338,
+      "loss": 4.8657,
+      "step": 4023
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017952502532131323,
+      "loss": 5.6647,
+      "step": 4024
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001796123137656398,
+      "loss": 3.8619,
+      "step": 4025
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017969943687415576,
+      "loss": 4.1589,
+      "step": 4026
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017978639446592718,
+      "loss": 4.6688,
+      "step": 4027
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017987318636036382,
+      "loss": 4.5705,
+      "step": 4028
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00017995981237721963,
+      "loss": 5.7671,
+      "step": 4029
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000180046272336593,
+      "loss": 5.1138,
+      "step": 4030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001801325660589271,
+      "loss": 3.7372,
+      "step": 4031
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018021869336501062,
+      "loss": 5.0526,
+      "step": 4032
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001803046540759775,
+      "loss": 4.3019,
+      "step": 4033
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001803904480133079,
+      "loss": 5.768,
+      "step": 4034
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018047607499882822,
+      "loss": 4.4483,
+      "step": 4035
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001805615348547117,
+      "loss": 4.7218,
+      "step": 4036
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018064682740347832,
+      "loss": 5.2206,
+      "step": 4037
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018073195246799602,
+      "loss": 4.7298,
+      "step": 4038
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018081690987148025,
+      "loss": 4.7248,
+      "step": 4039
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018090169943749476,
+      "loss": 4.3659,
+      "step": 4040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001809863209899517,
+      "loss": 3.9233,
+      "step": 4041
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018107077435311237,
+      "loss": 3.966,
+      "step": 4042
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018115505935158724,
+      "loss": 4.6424,
+      "step": 4043
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001812391758103364,
+      "loss": 3.9417,
+      "step": 4044
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018132312355467005,
+      "loss": 4.277,
+      "step": 4045
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018140690241024866,
+      "loss": 4.846,
+      "step": 4046
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018149051220308358,
+      "loss": 3.4144,
+      "step": 4047
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018157395275953718,
+      "loss": 5.1354,
+      "step": 4048
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018165722390632328,
+      "loss": 5.2445,
+      "step": 4049
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001817403254705075,
+      "loss": 4.0614,
+      "step": 4050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018182325727950777,
+      "loss": 4.1827,
+      "step": 4051
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018190601916109447,
+      "loss": 3.9915,
+      "step": 4052
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018198861094339077,
+      "loss": 4.3743,
+      "step": 4053
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018207103245487333,
+      "loss": 5.598,
+      "step": 4054
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018215328352437228,
+      "loss": 3.9864,
+      "step": 4055
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018223536398107176,
+      "loss": 4.8786,
+      "step": 4056
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018231727365451017,
+      "loss": 4.7117,
+      "step": 4057
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018239901237458064,
+      "loss": 5.6707,
+      "step": 4058
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018248057997153136,
+      "loss": 2.9685,
+      "step": 4059
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018256197627596583,
+      "loss": 3.917,
+      "step": 4060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001826432011188433,
+      "loss": 4.5489,
+      "step": 4061
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001827242543314791,
+      "loss": 2.6036,
+      "step": 4062
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001828051357455451,
+      "loss": 3.1723,
+      "step": 4063
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018288584519306984,
+      "loss": 4.5926,
+      "step": 4064
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000182966382506439,
+      "loss": 3.4394,
+      "step": 4065
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018304674751839576,
+      "loss": 3.95,
+      "step": 4066
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018312694006204119,
+      "loss": 5.5327,
+      "step": 4067
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018320695997083445,
+      "loss": 4.6194,
+      "step": 4068
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018328680707859317,
+      "loss": 3.2199,
+      "step": 4069
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018336648121949404,
+      "loss": 4.3732,
+      "step": 4070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018344598222807282,
+      "loss": 4.2959,
+      "step": 4071
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018352530993922481,
+      "loss": 4.9769,
+      "step": 4072
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018360446418820532,
+      "loss": 1.9274,
+      "step": 4073
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018368344481062979,
+      "loss": 4.5726,
+      "step": 4074
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018376225164247427,
+      "loss": 4.2226,
+      "step": 4075
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018384088452007578,
+      "loss": 2.369,
+      "step": 4076
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001839193432801325,
+      "loss": 3.8222,
+      "step": 4077
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018399762775970433,
+      "loss": 5.3088,
+      "step": 4078
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018407573779621306,
+      "loss": 5.4119,
+      "step": 4079
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018415367322744276,
+      "loss": 3.0444,
+      "step": 4080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018423143389154008,
+      "loss": 4.7797,
+      "step": 4081
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001843090196270146,
+      "loss": 5.6359,
+      "step": 4082
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018438643027273935,
+      "loss": 4.1936,
+      "step": 4083
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018446366566795077,
+      "loss": 3.4334,
+      "step": 4084
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001845407256522493,
+      "loss": 4.9421,
+      "step": 4085
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001846176100655998,
+      "loss": 2.7645,
+      "step": 4086
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018469431874833166,
+      "loss": 4.6274,
+      "step": 4087
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018477085154113912,
+      "loss": 4.8822,
+      "step": 4088
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018484720828508192,
+      "loss": 4.8649,
+      "step": 4089
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018492338882158523,
+      "loss": 3.6077,
+      "step": 4090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018499939299244025,
+      "loss": 3.6274,
+      "step": 4091
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018507522063980435,
+      "loss": 3.7645,
+      "step": 4092
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018515087160620168,
+      "loss": 1.9553,
+      "step": 4093
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001852263457345231,
+      "loss": 3.5519,
+      "step": 4094
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001853016428680269,
+      "loss": 3.7237,
+      "step": 4095
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018537676285033884,
+      "loss": 3.2056,
+      "step": 4096
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018545170552545265,
+      "loss": 5.9522,
+      "step": 4097
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018552647073773017,
+      "loss": 3.0478,
+      "step": 4098
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018560105833190194,
+      "loss": 4.8461,
+      "step": 4099
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018567546815306728,
+      "loss": 4.1119,
+      "step": 4100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018574970004669463,
+      "loss": 4.2586,
+      "step": 4101
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018582375385862215,
+      "loss": 4.0158,
+      "step": 4102
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018589762943505765,
+      "loss": 5.865,
+      "step": 4103
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001859713266225792,
+      "loss": 4.1522,
+      "step": 4104
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018604484526813523,
+      "loss": 5.2853,
+      "step": 4105
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018611818521904508,
+      "loss": 3.967,
+      "step": 4106
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018619134632299913,
+      "loss": 3.9026,
+      "step": 4107
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001862643284280592,
+      "loss": 4.2834,
+      "step": 4108
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018633713138265878,
+      "loss": 4.6061,
+      "step": 4109
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018640975503560355,
+      "loss": 4.7688,
+      "step": 4110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001864821992360715,
+      "loss": 5.1929,
+      "step": 4111
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018655446383361327,
+      "loss": 4.9148,
+      "step": 4112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018662654867815246,
+      "loss": 2.1447,
+      "step": 4113
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018669845361998604,
+      "loss": 4.5246,
+      "step": 4114
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018677017850978457,
+      "loss": 2.326,
+      "step": 4115
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018684172319859257,
+      "loss": 4.164,
+      "step": 4116
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018691308753782874,
+      "loss": 2.0505,
+      "step": 4117
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018698427137928632,
+      "loss": 5.6233,
+      "step": 4118
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001870552745751335,
+      "loss": 3.7954,
+      "step": 4119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018712609697791352,
+      "loss": 4.9987,
+      "step": 4120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018719673844054508,
+      "loss": 5.1215,
+      "step": 4121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018726719881632277,
+      "loss": 3.6172,
+      "step": 4122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018733747795891709,
+      "loss": 4.2318,
+      "step": 4123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018740757572237508,
+      "loss": 6.421,
+      "step": 4124
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018747749196112027,
+      "loss": 4.28,
+      "step": 4125
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018754722652995342,
+      "loss": 3.9631,
+      "step": 4126
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018761677928405242,
+      "loss": 3.4774,
+      "step": 4127
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001876861500789727,
+      "loss": 5.7775,
+      "step": 4128
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001877553387706477,
+      "loss": 4.132,
+      "step": 4129
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018782434521538895,
+      "loss": 3.3784,
+      "step": 4130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018789316926988653,
+      "loss": 2.6866,
+      "step": 4131
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001879618107912093,
+      "loss": 4.8302,
+      "step": 4132
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018803026963680507,
+      "loss": 4.7717,
+      "step": 4133
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018809854566450126,
+      "loss": 3.2786,
+      "step": 4134
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018816663873250474,
+      "loss": 4.6388,
+      "step": 4135
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018823454869940242,
+      "loss": 3.9513,
+      "step": 4136
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018830227542416153,
+      "loss": 4.2587,
+      "step": 4137
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001883698187661298,
+      "loss": 4.6158,
+      "step": 4138
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018843717858503571,
+      "loss": 5.7137,
+      "step": 4139
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018850435474098903,
+      "loss": 5.1516,
+      "step": 4140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001885713470944809,
+      "loss": 4.4785,
+      "step": 4141
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018863815550638412,
+      "loss": 4.6714,
+      "step": 4142
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018870477983795359,
+      "loss": 2.3869,
+      "step": 4143
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001887712199508264,
+      "loss": 3.8367,
+      "step": 4144
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018883747570702224,
+      "loss": 5.2524,
+      "step": 4145
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018890354696894372,
+      "loss": 4.7517,
+      "step": 4146
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018896943359937658,
+      "loss": 3.1402,
+      "step": 4147
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018903513546148997,
+      "loss": 3.7502,
+      "step": 4148
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001891006524188368,
+      "loss": 4.5174,
+      "step": 4149
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018916598433535388,
+      "loss": 4.8869,
+      "step": 4150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018923113107536244,
+      "loss": 3.992,
+      "step": 4151
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018929609250356827,
+      "loss": 3.2907,
+      "step": 4152
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018936086848506185,
+      "loss": 4.1865,
+      "step": 4153
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018942545888531897,
+      "loss": 2.9454,
+      "step": 4154
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018948986357020074,
+      "loss": 5.2073,
+      "step": 4155
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018955408240595395,
+      "loss": 6.6802,
+      "step": 4156
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018961811525921137,
+      "loss": 4.8292,
+      "step": 4157
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018968196199699205,
+      "loss": 3.4886,
+      "step": 4158
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018974562248670154,
+      "loss": 5.7377,
+      "step": 4159
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018980909659613215,
+      "loss": 5.3915,
+      "step": 4160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018987238419346326,
+      "loss": 5.2105,
+      "step": 4161
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018993548514726157,
+      "loss": 3.1397,
+      "step": 4162
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018999839932648156,
+      "loss": 2.6762,
+      "step": 4163
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019006112660046532,
+      "loss": 3.8416,
+      "step": 4164
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019012366683894333,
+      "loss": 4.0945,
+      "step": 4165
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001901860199120344,
+      "loss": 3.0966,
+      "step": 4166
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019024818569024602,
+      "loss": 5.7601,
+      "step": 4167
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001903101640444747,
+      "loss": 4.3446,
+      "step": 4168
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019037195484600616,
+      "loss": 3.1334,
+      "step": 4169
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019043355796651565,
+      "loss": 4.3705,
+      "step": 4170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019049497327806812,
+      "loss": 3.1011,
+      "step": 4171
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001905562006531186,
+      "loss": 3.1909,
+      "step": 4172
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019061723996451245,
+      "loss": 2.1465,
+      "step": 4173
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019067809108548554,
+      "loss": 3.6733,
+      "step": 4174
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019073875388966465,
+      "loss": 3.1926,
+      "step": 4175
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019079922825106747,
+      "loss": 4.8662,
+      "step": 4176
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019085951404410325,
+      "loss": 3.5566,
+      "step": 4177
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001909196111435727,
+      "loss": 3.2752,
+      "step": 4178
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001909795194246685,
+      "loss": 5.4414,
+      "step": 4179
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019103923876297545,
+      "loss": 5.381,
+      "step": 4180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019109876903447063,
+      "loss": 3.2592,
+      "step": 4181
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019115811011552388,
+      "loss": 4.2684,
+      "step": 4182
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019121726188289792,
+      "loss": 3.9145,
+      "step": 4183
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019127622421374866,
+      "loss": 4.337,
+      "step": 4184
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001913349969856253,
+      "loss": 4.5848,
+      "step": 4185
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019139358007647084,
+      "loss": 5.9864,
+      "step": 4186
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019145197336462217,
+      "loss": 3.4476,
+      "step": 4187
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019151017672881028,
+      "loss": 5.0302,
+      "step": 4188
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019156819004816073,
+      "loss": 2.9631,
+      "step": 4189
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019162601320219362,
+      "loss": 4.4382,
+      "step": 4190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019168364607082406,
+      "loss": 3.5049,
+      "step": 4191
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019174108853436232,
+      "loss": 3.0574,
+      "step": 4192
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019179834047351404,
+      "loss": 3.3025,
+      "step": 4193
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019185540176938062,
+      "loss": 6.0765,
+      "step": 4194
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019191227230345935,
+      "loss": 3.2504,
+      "step": 4195
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019196895195764362,
+      "loss": 4.3593,
+      "step": 4196
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019202544061422335,
+      "loss": 4.3057,
+      "step": 4197
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019208173815588503,
+      "loss": 4.8534,
+      "step": 4198
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019213784446571209,
+      "loss": 3.8868,
+      "step": 4199
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019219375942718507,
+      "loss": 3.0987,
+      "step": 4200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019224948292418195,
+      "loss": 4.1861,
+      "step": 4201
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019230501484097828,
+      "loss": 5.1527,
+      "step": 4202
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019236035506224746,
+      "loss": 3.2142,
+      "step": 4203
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019241550347306105,
+      "loss": 3.5671,
+      "step": 4204
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000192470459958889,
+      "loss": 4.1947,
+      "step": 4205
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019252522440559974,
+      "loss": 5.0718,
+      "step": 4206
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019257979669946057,
+      "loss": 3.632,
+      "step": 4207
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001926341767271378,
+      "loss": 3.5481,
+      "step": 4208
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019268836437569712,
+      "loss": 4.2741,
+      "step": 4209
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019274235953260365,
+      "loss": 4.3293,
+      "step": 4210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019279616208572233,
+      "loss": 3.4052,
+      "step": 4211
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019284977192331803,
+      "loss": 5.8669,
+      "step": 4212
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019290318893405597,
+      "loss": 4.2041,
+      "step": 4213
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001929564130070017,
+      "loss": 4.4892,
+      "step": 4214
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019300944403162148,
+      "loss": 3.5933,
+      "step": 4215
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019306228189778254,
+      "loss": 5.5635,
+      "step": 4216
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001931149264957532,
+      "loss": 3.8591,
+      "step": 4217
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019316737771620318,
+      "loss": 3.4884,
+      "step": 4218
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019321963545020378,
+      "loss": 3.5044,
+      "step": 4219
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019327169958922814,
+      "loss": 3.587,
+      "step": 4220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019332357002515142,
+      "loss": 3.8158,
+      "step": 4221
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019337524665025116,
+      "loss": 3.1889,
+      "step": 4222
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001934267293572072,
+      "loss": 2.3629,
+      "step": 4223
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019347801803910235,
+      "loss": 3.5379,
+      "step": 4224
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019352911258942217,
+      "loss": 3.8719,
+      "step": 4225
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019358001290205543,
+      "loss": 5.0306,
+      "step": 4226
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019363071887129433,
+      "loss": 2.8466,
+      "step": 4227
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019368123039183467,
+      "loss": 4.1445,
+      "step": 4228
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019373154735877603,
+      "loss": 5.7359,
+      "step": 4229
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019378166966762205,
+      "loss": 4.9637,
+      "step": 4230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019383159721428065,
+      "loss": 5.5845,
+      "step": 4231
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019388132989506422,
+      "loss": 3.6032,
+      "step": 4232
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019393086760668984,
+      "loss": 4.9698,
+      "step": 4233
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001939802102462794,
+      "loss": 5.1821,
+      "step": 4234
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001940293577113601,
+      "loss": 4.3389,
+      "step": 4235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001940783098998643,
+      "loss": 4.9597,
+      "step": 4236
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019412706671012995,
+      "loss": 4.1264,
+      "step": 4237
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019417562804090082,
+      "loss": 3.8933,
+      "step": 4238
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001942239937913266,
+      "loss": 5.7195,
+      "step": 4239
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001942721638609631,
+      "loss": 4.5997,
+      "step": 4240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019432013814977257,
+      "loss": 4.0519,
+      "step": 4241
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019436791655812386,
+      "loss": 4.5996,
+      "step": 4242
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019441549898679261,
+      "loss": 4.4596,
+      "step": 4243
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001944628853369614,
+      "loss": 3.0365,
+      "step": 4244
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019451007551022013,
+      "loss": 4.5907,
+      "step": 4245
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000194557069408566,
+      "loss": 4.2887,
+      "step": 4246
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019460386693440394,
+      "loss": 3.5658,
+      "step": 4247
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019465046799054656,
+      "loss": 3.9146,
+      "step": 4248
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001946968724802146,
+      "loss": 4.7739,
+      "step": 4249
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000194743080307037,
+      "loss": 4.7124,
+      "step": 4250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019478909137505112,
+      "loss": 5.2874,
+      "step": 4251
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001948349055887029,
+      "loss": 4.8943,
+      "step": 4252
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019488052285284715,
+      "loss": 3.5171,
+      "step": 4253
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019492594307274764,
+      "loss": 5.2672,
+      "step": 4254
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019497116615407748,
+      "loss": 4.5248,
+      "step": 4255
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019501619200291908,
+      "loss": 4.8094,
+      "step": 4256
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001950610205257644,
+      "loss": 4.35,
+      "step": 4257
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019510565162951537,
+      "loss": 4.0407,
+      "step": 4258
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019515008522148376,
+      "loss": 3.9306,
+      "step": 4259
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001951943212093916,
+      "loss": 5.1323,
+      "step": 4260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019523835950137128,
+      "loss": 5.2756,
+      "step": 4261
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019528220000596575,
+      "loss": 5.8998,
+      "step": 4262
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019532584263212878,
+      "loss": 3.8926,
+      "step": 4263
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019536928728922494,
+      "loss": 4.7362,
+      "step": 4264
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001954125338870301,
+      "loss": 4.5688,
+      "step": 4265
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019545558233573134,
+      "loss": 2.6199,
+      "step": 4266
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019549843254592733,
+      "loss": 5.0206,
+      "step": 4267
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019554108442862832,
+      "loss": 2.9526,
+      "step": 4268
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019558353789525662,
+      "loss": 3.3843,
+      "step": 4269
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019562579285764637,
+      "loss": 3.8948,
+      "step": 4270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019566784922804415,
+      "loss": 3.6174,
+      "step": 4271
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001957097069191089,
+      "loss": 4.1019,
+      "step": 4272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001957513658439121,
+      "loss": 4.5572,
+      "step": 4273
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001957928259159382,
+      "loss": 5.3222,
+      "step": 4274
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019583408704908442,
+      "loss": 4.4712,
+      "step": 4275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019587514915766124,
+      "loss": 4.8489,
+      "step": 4276
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001959160121563924,
+      "loss": 3.7989,
+      "step": 4277
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019595667596041528,
+      "loss": 2.6331,
+      "step": 4278
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019599714048528077,
+      "loss": 5.0813,
+      "step": 4279
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019603740564695375,
+      "loss": 2.5456,
+      "step": 4280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019607747136181298,
+      "loss": 3.5177,
+      "step": 4281
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001961173375466516,
+      "loss": 5.0204,
+      "step": 4282
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019615700411867697,
+      "loss": 4.4016,
+      "step": 4283
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019619647099551114,
+      "loss": 3.5587,
+      "step": 4284
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019623573809519077,
+      "loss": 5.8158,
+      "step": 4285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019627480533616746,
+      "loss": 4.8404,
+      "step": 4286
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019631367263730787,
+      "loss": 4.1807,
+      "step": 4287
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019635233991789386,
+      "loss": 4.96,
+      "step": 4288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019639080709762268,
+      "loss": 3.852,
+      "step": 4289
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019642907409660719,
+      "loss": 3.4305,
+      "step": 4290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019646714083537594,
+      "loss": 6.0307,
+      "step": 4291
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019650500723487335,
+      "loss": 4.2654,
+      "step": 4292
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019654267321646,
+      "loss": 4.1548,
+      "step": 4293
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019658013870191258,
+      "loss": 3.642,
+      "step": 4294
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019661740361342415,
+      "loss": 3.6638,
+      "step": 4295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001966544678736044,
+      "loss": 6.1919,
+      "step": 4296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019669133140547978,
+      "loss": 4.1255,
+      "step": 4297
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001967279941324934,
+      "loss": 4.9473,
+      "step": 4298
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019676445597850554,
+      "loss": 3.1888,
+      "step": 4299
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019680071686779366,
+      "loss": 4.9627,
+      "step": 4300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019683677672505253,
+      "loss": 5.4431,
+      "step": 4301
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019687263547539437,
+      "loss": 4.1322,
+      "step": 4302
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019690829304434916,
+      "loss": 4.5497,
+      "step": 4303
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019694374935786457,
+      "loss": 4.6009,
+      "step": 4304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019697900434230629,
+      "loss": 4.5276,
+      "step": 4305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019701405792445814,
+      "loss": 3.522,
+      "step": 4306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019704891003152216,
+      "loss": 3.4753,
+      "step": 4307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001970835605911189,
+      "loss": 4.5345,
+      "step": 4308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001971180095312873,
+      "loss": 3.2991,
+      "step": 4309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019715225678048514,
+      "loss": 4.4241,
+      "step": 4310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019718630226758907,
+      "loss": 4.5228,
+      "step": 4311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001972201459218947,
+      "loss": 4.4727,
+      "step": 4312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019725378767311685,
+      "loss": 5.0401,
+      "step": 4313
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001972872274513896,
+      "loss": 4.8386,
+      "step": 4314
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001973204651872665,
+      "loss": 5.2224,
+      "step": 4315
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019735350081172067,
+      "loss": 4.111,
+      "step": 4316
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000197386334256145,
+      "loss": 3.6074,
+      "step": 4317
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019741896545235227,
+      "loss": 3.7133,
+      "step": 4318
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019745139433257526,
+      "loss": 4.226,
+      "step": 4319
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019748362082946686,
+      "loss": 4.8672,
+      "step": 4320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019751564487610038,
+      "loss": 3.6439,
+      "step": 4321
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019754746640596944,
+      "loss": 5.3177,
+      "step": 4322
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019757908535298832,
+      "loss": 4.0092,
+      "step": 4323
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019761050165149208,
+      "loss": 3.0635,
+      "step": 4324
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001976417152362364,
+      "loss": 4.8065,
+      "step": 4325
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019767272604239824,
+      "loss": 1.9274,
+      "step": 4326
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019770353400557546,
+      "loss": 3.3433,
+      "step": 4327
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001977341390617873,
+      "loss": 4.6842,
+      "step": 4328
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001977645411474743,
+      "loss": 4.1791,
+      "step": 4329
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019779474019949862,
+      "loss": 4.3212,
+      "step": 4330
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000197824736155144,
+      "loss": 4.8481,
+      "step": 4331
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019785452895211604,
+      "loss": 3.7019,
+      "step": 4332
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019788411852854213,
+      "loss": 3.8597,
+      "step": 4333
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019791350482297178,
+      "loss": 2.2169,
+      "step": 4334
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019794268777437667,
+      "loss": 2.3188,
+      "step": 4335
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019797166732215076,
+      "loss": 3.9671,
+      "step": 4336
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001980004434061104,
+      "loss": 4.5561,
+      "step": 4337
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001980290159664946,
+      "loss": 4.8317,
+      "step": 4338
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019805738494396488,
+      "loss": 3.938,
+      "step": 4339
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019808555027960568,
+      "loss": 6.4169,
+      "step": 4340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001981135119149243,
+      "loss": 5.7964,
+      "step": 4341
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019814126979185104,
+      "loss": 4.1002,
+      "step": 4342
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019816882385273945,
+      "loss": 1.8383,
+      "step": 4343
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019819617404036628,
+      "loss": 5.2311,
+      "step": 4344
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019822332029793173,
+      "loss": 3.5985,
+      "step": 4345
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019825026256905948,
+      "loss": 3.4469,
+      "step": 4346
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019827700079779682,
+      "loss": 5.6609,
+      "step": 4347
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019830353492861487,
+      "loss": 4.7354,
+      "step": 4348
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019832986490640856,
+      "loss": 4.9683,
+      "step": 4349
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019835599067649678,
+      "loss": 5.3011,
+      "step": 4350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001983819121846225,
+      "loss": 3.1021,
+      "step": 4351
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019840762937695295,
+      "loss": 4.6209,
+      "step": 4352
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019843314220007968,
+      "loss": 5.4262,
+      "step": 4353
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019845845060101856,
+      "loss": 5.498,
+      "step": 4354
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019848355452721012,
+      "loss": 3.767,
+      "step": 4355
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019850845392651952,
+      "loss": 5.8622,
+      "step": 4356
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019853314874723657,
+      "loss": 5.6982,
+      "step": 4357
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000198557638938076,
+      "loss": 4.5925,
+      "step": 4358
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001985819244481776,
+      "loss": 4.3072,
+      "step": 4359
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019860600522710612,
+      "loss": 4.4345,
+      "step": 4360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019862988122485144,
+      "loss": 5.7037,
+      "step": 4361
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001986535523918289,
+      "loss": 2.9157,
+      "step": 4362
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019867701867887908,
+      "loss": 6.5688,
+      "step": 4363
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001987002800372681,
+      "loss": 4.9866,
+      "step": 4364
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019872333641868768,
+      "loss": 3.4773,
+      "step": 4365
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019874618777525517,
+      "loss": 4.3187,
+      "step": 4366
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019876883405951377,
+      "loss": 4.291,
+      "step": 4367
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019879127522443255,
+      "loss": 4.1,
+      "step": 4368
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019881351122340654,
+      "loss": 3.6076,
+      "step": 4369
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019883554201025687,
+      "loss": 3.337,
+      "step": 4370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001988573675392308,
+      "loss": 3.5263,
+      "step": 4371
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019887898776500202,
+      "loss": 2.9075,
+      "step": 4372
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019890040264267035,
+      "loss": 3.6854,
+      "step": 4373
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001989216121277623,
+      "loss": 4.3511,
+      "step": 4374
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019894261617623072,
+      "loss": 3.5308,
+      "step": 4375
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019896341474445525,
+      "loss": 4.928,
+      "step": 4376
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001989840077892422,
+      "loss": 6.116,
+      "step": 4377
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019900439526782479,
+      "loss": 4.8613,
+      "step": 4378
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000199024577137863,
+      "loss": 2.7898,
+      "step": 4379
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019904455335744393,
+      "loss": 3.1813,
+      "step": 4380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019906432388508175,
+      "loss": 4.1474,
+      "step": 4381
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019908388867971772,
+      "loss": 3.8655,
+      "step": 4382
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019910324770072048,
+      "loss": 4.1452,
+      "step": 4383
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019912240090788595,
+      "loss": 5.0234,
+      "step": 4384
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019914134826143747,
+      "loss": 4.1831,
+      "step": 4385
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019916008972202585,
+      "loss": 4.2842,
+      "step": 4386
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019917862525072966,
+      "loss": 4.916,
+      "step": 4387
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001991969548090549,
+      "loss": 2.1136,
+      "step": 4388
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019921507835893553,
+      "loss": 5.1165,
+      "step": 4389
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019923299586273324,
+      "loss": 2.7861,
+      "step": 4390
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019925070728323758,
+      "loss": 5.0809,
+      "step": 4391
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019926821258366622,
+      "loss": 2.8444,
+      "step": 4392
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019928551172766477,
+      "loss": 2.8434,
+      "step": 4393
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019930260467930706,
+      "loss": 4.7674,
+      "step": 4394
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001993194914030951,
+      "loss": 4.5014,
+      "step": 4395
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019933617186395917,
+      "loss": 2.8873,
+      "step": 4396
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019935264602725795,
+      "loss": 4.2748,
+      "step": 4397
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001993689138587785,
+      "loss": 3.7597,
+      "step": 4398
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019938497532473646,
+      "loss": 3.9001,
+      "step": 4399
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001994008303917759,
+      "loss": 3.9354,
+      "step": 4400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019941647902696973,
+      "loss": 3.2311,
+      "step": 4401
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019943192119781938,
+      "loss": 4.7518,
+      "step": 4402
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019944715687225518,
+      "loss": 5.5789,
+      "step": 4403
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019946218601863625,
+      "loss": 4.8766,
+      "step": 4404
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019947700860575065,
+      "loss": 4.5741,
+      "step": 4405
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001994916246028154,
+      "loss": 4.3871,
+      "step": 4406
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001995060339794765,
+      "loss": 4.4344,
+      "step": 4407
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019952023670580918,
+      "loss": 3.2224,
+      "step": 4408
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001995342327523177,
+      "loss": 2.8095,
+      "step": 4409
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019954802208993564,
+      "loss": 4.5132,
+      "step": 4410
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019956160469002584,
+      "loss": 4.7276,
+      "step": 4411
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019957498052438046,
+      "loss": 4.9196,
+      "step": 4412
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019958814956522108,
+      "loss": 4.7696,
+      "step": 4413
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001996011117851987,
+      "loss": 4.7389,
+      "step": 4414
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019961386715739395,
+      "loss": 3.6579,
+      "step": 4415
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019962641565531692,
+      "loss": 5.2974,
+      "step": 4416
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019963875725290738,
+      "loss": 3.9383,
+      "step": 4417
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019965089192453475,
+      "loss": 3.0198,
+      "step": 4418
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019966281964499825,
+      "loss": 4.1165,
+      "step": 4419
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019967454038952677,
+      "loss": 3.1906,
+      "step": 4420
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019968605413377925,
+      "loss": 4.2974,
+      "step": 4421
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001996973608538442,
+      "loss": 6.8364,
+      "step": 4422
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019970846052624043,
+      "loss": 3.2336,
+      "step": 4423
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019971935312791645,
+      "loss": 5.1803,
+      "step": 4424
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000199730038636251,
+      "loss": 5.6177,
+      "step": 4425
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019974051702905277,
+      "loss": 5.5883,
+      "step": 4426
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019975078828456065,
+      "loss": 6.18,
+      "step": 4427
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001997608523814437,
+      "loss": 4.7102,
+      "step": 4428
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019977070929880114,
+      "loss": 5.2304,
+      "step": 4429
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001997803590161626,
+      "loss": 4.4427,
+      "step": 4430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019978980151348782,
+      "loss": 4.275,
+      "step": 4431
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019979903677116702,
+      "loss": 2.749,
+      "step": 4432
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019980806477002078,
+      "loss": 4.0765,
+      "step": 4433
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019981688549130006,
+      "loss": 3.7824,
+      "step": 4434
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019982549891668639,
+      "loss": 4.0541,
+      "step": 4435
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019983390502829166,
+      "loss": 4.0262,
+      "step": 4436
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001998421038086584,
+      "loss": 4.849,
+      "step": 4437
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019985009524075978,
+      "loss": 4.2866,
+      "step": 4438
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019985787930799939,
+      "loss": 4.2617,
+      "step": 4439
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001998654559942116,
+      "loss": 4.291,
+      "step": 4440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019987282528366148,
+      "loss": 3.2615,
+      "step": 4441
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019987998716104472,
+      "loss": 3.6954,
+      "step": 4442
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019988694161148787,
+      "loss": 4.4301,
+      "step": 4443
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019989368862054812,
+      "loss": 4.1962,
+      "step": 4444
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019990022817421358,
+      "loss": 3.0085,
+      "step": 4445
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019990656025890315,
+      "loss": 3.6284,
+      "step": 4446
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019991268486146662,
+      "loss": 4.0809,
+      "step": 4447
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999186019691846,
+      "loss": 4.0912,
+      "step": 4448
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999243115697687,
+      "loss": 4.0246,
+      "step": 4449
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019992981365136143,
+      "loss": 2.3051,
+      "step": 4450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999351082025363,
+      "loss": 3.3251,
+      "step": 4451
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019994019521229776,
+      "loss": 4.8436,
+      "step": 4452
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019994507467008133,
+      "loss": 2.1878,
+      "step": 4453
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999497465657535,
+      "loss": 1.8606,
+      "step": 4454
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999542108896119,
+      "loss": 4.8264,
+      "step": 4455
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019995846763238512,
+      "loss": 5.106,
+      "step": 4456
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000199962516785233,
+      "loss": 4.3687,
+      "step": 4457
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019996635833974637,
+      "loss": 5.316,
+      "step": 4458
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019996999228794727,
+      "loss": 4.3448,
+      "step": 4459
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019997341862228884,
+      "loss": 5.0116,
+      "step": 4460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019997663733565537,
+      "loss": 2.23,
+      "step": 4461
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999796484213624,
+      "loss": 4.0596,
+      "step": 4462
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019998245187315662,
+      "loss": 4.5042,
+      "step": 4463
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019998504768521587,
+      "loss": 5.3704,
+      "step": 4464
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019998743585214932,
+      "loss": 4.5394,
+      "step": 4465
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019998961636899734,
+      "loss": 4.6123,
+      "step": 4466
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999158923123144,
+      "loss": 2.3017,
+      "step": 4467
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999335443475453,
+      "loss": 4.9044,
+      "step": 4468
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999491197590066,
+      "loss": 5.0815,
+      "step": 4469
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999626185143518,
+      "loss": 5.5885,
+      "step": 4470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999740405855478,
+      "loss": 4.65,
+      "step": 4471
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999983385948873,
+      "loss": 3.4558,
+      "step": 4472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999906545849197,
+      "loss": 5.2408,
+      "step": 4473
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999958464785923,
+      "loss": 4.3937,
+      "step": 4474
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999998961619109,
+      "loss": 3.9685,
+      "step": 4475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002,
+      "loss": 5.8921,
+      "step": 4476
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999998961619109,
+      "loss": 3.6178,
+      "step": 4477
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999958464785923,
+      "loss": 4.0317,
+      "step": 4478
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999906545849197,
+      "loss": 4.3238,
+      "step": 4479
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999983385948873,
+      "loss": 4.8197,
+      "step": 4480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999740405855478,
+      "loss": 4.2103,
+      "step": 4481
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999626185143518,
+      "loss": 3.241,
+      "step": 4482
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999491197590066,
+      "loss": 4.0354,
+      "step": 4483
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999335443475453,
+      "loss": 4.9373,
+      "step": 4484
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019999158923123144,
+      "loss": 4.7876,
+      "step": 4485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019998961636899734,
+      "loss": 4.2537,
+      "step": 4486
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019998743585214932,
+      "loss": 5.0783,
+      "step": 4487
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019998504768521587,
+      "loss": 6.1064,
+      "step": 4488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019998245187315662,
+      "loss": 3.4014,
+      "step": 4489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001999796484213624,
+      "loss": 3.761,
+      "step": 4490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019997663733565537,
+      "loss": 5.0864,
+      "step": 4491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019997341862228884,
+      "loss": 4.1403,
+      "step": 4492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019996999228794727,
+      "loss": 3.395,
+      "step": 4493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019996635833974637,
+      "loss": 4.1175,
+      "step": 4494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000199962516785233,
+      "loss": 3.5777,
+      "step": 4495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019995846763238512,
+      "loss": 6.3329,
+      "step": 4496
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001999542108896119,
+      "loss": 6.3208,
+      "step": 4497
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001999497465657535,
+      "loss": 4.9116,
+      "step": 4498
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019994507467008133,
+      "loss": 4.5808,
+      "step": 4499
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019994019521229776,
+      "loss": 3.0961,
+      "step": 4500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001999351082025363,
+      "loss": 3.4385,
+      "step": 4501
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019992981365136143,
+      "loss": 3.9727,
+      "step": 4502
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001999243115697687,
+      "loss": 2.7536,
+      "step": 4503
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001999186019691846,
+      "loss": 4.1242,
+      "step": 4504
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019991268486146662,
+      "loss": 4.1156,
+      "step": 4505
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019990656025890315,
+      "loss": 3.8481,
+      "step": 4506
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019990022817421358,
+      "loss": 4.7274,
+      "step": 4507
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019989368862054812,
+      "loss": 4.3972,
+      "step": 4508
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019988694161148787,
+      "loss": 2.4545,
+      "step": 4509
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019987998716104472,
+      "loss": 4.228,
+      "step": 4510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019987282528366148,
+      "loss": 5.6057,
+      "step": 4511
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001998654559942116,
+      "loss": 5.4119,
+      "step": 4512
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019985787930799939,
+      "loss": 4.283,
+      "step": 4513
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019985009524075978,
+      "loss": 4.3171,
+      "step": 4514
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019984210380865844,
+      "loss": 5.7483,
+      "step": 4515
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019983390502829166,
+      "loss": 5.1826,
+      "step": 4516
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019982549891668639,
+      "loss": 5.7175,
+      "step": 4517
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019981688549130006,
+      "loss": 5.0499,
+      "step": 4518
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019980806477002078,
+      "loss": 5.76,
+      "step": 4519
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019979903677116702,
+      "loss": 5.2731,
+      "step": 4520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019978980151348782,
+      "loss": 4.7289,
+      "step": 4521
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001997803590161626,
+      "loss": 3.3798,
+      "step": 4522
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019977070929880116,
+      "loss": 5.2759,
+      "step": 4523
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001997608523814437,
+      "loss": 4.2782,
+      "step": 4524
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019975078828456065,
+      "loss": 4.4148,
+      "step": 4525
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019974051702905277,
+      "loss": 2.6902,
+      "step": 4526
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000199730038636251,
+      "loss": 2.0607,
+      "step": 4527
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019971935312791645,
+      "loss": 2.3536,
+      "step": 4528
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019970846052624043,
+      "loss": 3.1263,
+      "step": 4529
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001996973608538442,
+      "loss": 4.3599,
+      "step": 4530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019968605413377922,
+      "loss": 3.7734,
+      "step": 4531
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019967454038952677,
+      "loss": 4.508,
+      "step": 4532
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019966281964499825,
+      "loss": 2.4569,
+      "step": 4533
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019965089192453475,
+      "loss": 4.7286,
+      "step": 4534
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019963875725290738,
+      "loss": 3.2577,
+      "step": 4535
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019962641565531692,
+      "loss": 5.5859,
+      "step": 4536
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019961386715739395,
+      "loss": 3.039,
+      "step": 4537
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001996011117851987,
+      "loss": 3.1585,
+      "step": 4538
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019958814956522108,
+      "loss": 4.9201,
+      "step": 4539
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019957498052438046,
+      "loss": 4.6341,
+      "step": 4540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019956160469002584,
+      "loss": 5.5972,
+      "step": 4541
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019954802208993564,
+      "loss": 4.7859,
+      "step": 4542
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001995342327523177,
+      "loss": 4.3141,
+      "step": 4543
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019952023670580918,
+      "loss": 4.3946,
+      "step": 4544
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001995060339794765,
+      "loss": 4.166,
+      "step": 4545
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001994916246028154,
+      "loss": 3.5269,
+      "step": 4546
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019947700860575065,
+      "loss": 5.3439,
+      "step": 4547
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019946218601863625,
+      "loss": 5.6791,
+      "step": 4548
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019944715687225518,
+      "loss": 5.3832,
+      "step": 4549
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019943192119781938,
+      "loss": 3.9807,
+      "step": 4550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019941647902696973,
+      "loss": 4.9374,
+      "step": 4551
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019940083039177593,
+      "loss": 4.3012,
+      "step": 4552
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019938497532473646,
+      "loss": 4.8701,
+      "step": 4553
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001993689138587785,
+      "loss": 4.2699,
+      "step": 4554
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019935264602725797,
+      "loss": 3.7356,
+      "step": 4555
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001993361718639592,
+      "loss": 2.63,
+      "step": 4556
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001993194914030951,
+      "loss": 4.6151,
+      "step": 4557
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019930260467930706,
+      "loss": 3.2242,
+      "step": 4558
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019928551172766477,
+      "loss": 5.4046,
+      "step": 4559
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001992682125836662,
+      "loss": 3.3992,
+      "step": 4560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019925070728323758,
+      "loss": 4.8683,
+      "step": 4561
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019923299586273324,
+      "loss": 3.0179,
+      "step": 4562
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019921507835893553,
+      "loss": 3.3169,
+      "step": 4563
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001991969548090549,
+      "loss": 5.3046,
+      "step": 4564
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019917862525072966,
+      "loss": 3.6705,
+      "step": 4565
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019916008972202585,
+      "loss": 4.0912,
+      "step": 4566
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019914134826143747,
+      "loss": 4.4508,
+      "step": 4567
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019912240090788595,
+      "loss": 3.9661,
+      "step": 4568
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001991032477007205,
+      "loss": 3.2954,
+      "step": 4569
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019908388867971772,
+      "loss": 5.3812,
+      "step": 4570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019906432388508175,
+      "loss": 4.6724,
+      "step": 4571
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019904455335744393,
+      "loss": 4.3605,
+      "step": 4572
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000199024577137863,
+      "loss": 3.5069,
+      "step": 4573
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019900439526782479,
+      "loss": 4.4659,
+      "step": 4574
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001989840077892422,
+      "loss": 4.8092,
+      "step": 4575
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019896341474445528,
+      "loss": 4.4295,
+      "step": 4576
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019894261617623072,
+      "loss": 3.2486,
+      "step": 4577
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001989216121277623,
+      "loss": 2.5837,
+      "step": 4578
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019890040264267038,
+      "loss": 4.5619,
+      "step": 4579
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019887898776500205,
+      "loss": 4.6038,
+      "step": 4580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019885736753923086,
+      "loss": 4.6206,
+      "step": 4581
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019883554201025687,
+      "loss": 4.2231,
+      "step": 4582
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019881351122340654,
+      "loss": 2.7153,
+      "step": 4583
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019879127522443257,
+      "loss": 3.5026,
+      "step": 4584
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019876883405951377,
+      "loss": 3.0793,
+      "step": 4585
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001987461877752552,
+      "loss": 4.0067,
+      "step": 4586
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001987233364186877,
+      "loss": 3.3168,
+      "step": 4587
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019870028003726812,
+      "loss": 5.1679,
+      "step": 4588
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019867701867887908,
+      "loss": 4.6201,
+      "step": 4589
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001986535523918289,
+      "loss": 1.924,
+      "step": 4590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019862988122485144,
+      "loss": 4.2516,
+      "step": 4591
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001986060052271061,
+      "loss": 3.9165,
+      "step": 4592
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001985819244481776,
+      "loss": 2.6062,
+      "step": 4593
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000198557638938076,
+      "loss": 5.0059,
+      "step": 4594
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019853314874723654,
+      "loss": 4.3218,
+      "step": 4595
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001985084539265195,
+      "loss": 4.2153,
+      "step": 4596
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019848355452721012,
+      "loss": 3.4939,
+      "step": 4597
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019845845060101856,
+      "loss": 4.3029,
+      "step": 4598
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019843314220007968,
+      "loss": 2.3983,
+      "step": 4599
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019840762937695295,
+      "loss": 3.4054,
+      "step": 4600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001983819121846225,
+      "loss": 4.9394,
+      "step": 4601
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019835599067649678,
+      "loss": 5.6835,
+      "step": 4602
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019832986490640856,
+      "loss": 4.5381,
+      "step": 4603
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001983035349286149,
+      "loss": 2.3902,
+      "step": 4604
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019827700079779685,
+      "loss": 4.0048,
+      "step": 4605
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019825026256905948,
+      "loss": 5.1497,
+      "step": 4606
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019822332029793173,
+      "loss": 5.1424,
+      "step": 4607
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981961740403663,
+      "loss": 4.7435,
+      "step": 4608
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019816882385273945,
+      "loss": 4.2428,
+      "step": 4609
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019814126979185104,
+      "loss": 3.2768,
+      "step": 4610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981135119149243,
+      "loss": 3.6423,
+      "step": 4611
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019808555027960568,
+      "loss": 4.2128,
+      "step": 4612
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001980573849439649,
+      "loss": 3.2155,
+      "step": 4613
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001980290159664946,
+      "loss": 4.87,
+      "step": 4614
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019800044340611043,
+      "loss": 4.0221,
+      "step": 4615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001979716673221508,
+      "loss": 3.8939,
+      "step": 4616
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019794268777437667,
+      "loss": 2.9579,
+      "step": 4617
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019791350482297178,
+      "loss": 5.5659,
+      "step": 4618
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019788411852854213,
+      "loss": 5.337,
+      "step": 4619
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019785452895211604,
+      "loss": 3.4716,
+      "step": 4620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019782473615514404,
+      "loss": 3.8185,
+      "step": 4621
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019779474019949862,
+      "loss": 4.7107,
+      "step": 4622
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001977645411474743,
+      "loss": 5.4157,
+      "step": 4623
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001977341390617873,
+      "loss": 3.8622,
+      "step": 4624
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019770353400557544,
+      "loss": 5.5508,
+      "step": 4625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019767272604239821,
+      "loss": 3.1076,
+      "step": 4626
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001976417152362364,
+      "loss": 5.006,
+      "step": 4627
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019761050165149205,
+      "loss": 5.2942,
+      "step": 4628
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019757908535298832,
+      "loss": 3.2217,
+      "step": 4629
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019754746640596947,
+      "loss": 2.3791,
+      "step": 4630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019751564487610038,
+      "loss": 2.9265,
+      "step": 4631
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001974836208294669,
+      "loss": 5.1747,
+      "step": 4632
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019745139433257526,
+      "loss": 4.0901,
+      "step": 4633
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001974189654523523,
+      "loss": 3.8448,
+      "step": 4634
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000197386334256145,
+      "loss": 5.5707,
+      "step": 4635
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019735350081172067,
+      "loss": 2.6986,
+      "step": 4636
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001973204651872665,
+      "loss": 5.2276,
+      "step": 4637
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001972872274513896,
+      "loss": 4.8918,
+      "step": 4638
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019725378767311685,
+      "loss": 3.6544,
+      "step": 4639
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019722014592189472,
+      "loss": 3.3061,
+      "step": 4640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001971863022675891,
+      "loss": 5.1107,
+      "step": 4641
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019715225678048514,
+      "loss": 5.1161,
+      "step": 4642
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001971180095312873,
+      "loss": 5.2023,
+      "step": 4643
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001970835605911189,
+      "loss": 3.3528,
+      "step": 4644
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001970489100315222,
+      "loss": 4.9715,
+      "step": 4645
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019701405792445817,
+      "loss": 3.6021,
+      "step": 4646
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001969790043423063,
+      "loss": 3.5893,
+      "step": 4647
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001969437493578646,
+      "loss": 2.7662,
+      "step": 4648
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019690829304434918,
+      "loss": 1.0966,
+      "step": 4649
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001968726354753944,
+      "loss": 3.6514,
+      "step": 4650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019683677672505256,
+      "loss": 5.3157,
+      "step": 4651
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001968007168677937,
+      "loss": 5.0594,
+      "step": 4652
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019676445597850557,
+      "loss": 3.8083,
+      "step": 4653
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001967279941324934,
+      "loss": 5.3882,
+      "step": 4654
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019669133140547978,
+      "loss": 4.4977,
+      "step": 4655
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001966544678736044,
+      "loss": 3.3027,
+      "step": 4656
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019661740361342413,
+      "loss": 4.9503,
+      "step": 4657
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019658013870191255,
+      "loss": 4.9524,
+      "step": 4658
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019654267321645997,
+      "loss": 4.0133,
+      "step": 4659
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019650500723487335,
+      "loss": 4.135,
+      "step": 4660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019646714083537594,
+      "loss": 3.9839,
+      "step": 4661
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001964290740966072,
+      "loss": 4.43,
+      "step": 4662
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019639080709762268,
+      "loss": 4.6458,
+      "step": 4663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001963523399178939,
+      "loss": 4.3038,
+      "step": 4664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001963136726373079,
+      "loss": 4.6532,
+      "step": 4665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019627480533616752,
+      "loss": 4.4089,
+      "step": 4666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001962357380951908,
+      "loss": 3.9896,
+      "step": 4667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019619647099551116,
+      "loss": 3.9399,
+      "step": 4668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019615700411867697,
+      "loss": 4.3282,
+      "step": 4669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001961173375466516,
+      "loss": 3.842,
+      "step": 4670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019607747136181298,
+      "loss": 3.1345,
+      "step": 4671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019603740564695375,
+      "loss": 3.3394,
+      "step": 4672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001959971404852808,
+      "loss": 4.5207,
+      "step": 4673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001959566759604153,
+      "loss": 5.4793,
+      "step": 4674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019591601215639243,
+      "loss": 4.6574,
+      "step": 4675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019587514915766124,
+      "loss": 4.7794,
+      "step": 4676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019583408704908445,
+      "loss": 4.2273,
+      "step": 4677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001957928259159382,
+      "loss": 3.9085,
+      "step": 4678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019575136584391216,
+      "loss": 3.8977,
+      "step": 4679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019570970691910894,
+      "loss": 3.2787,
+      "step": 4680
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001956678492280442,
+      "loss": 3.4111,
+      "step": 4681
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019562579285764642,
+      "loss": 4.2865,
+      "step": 4682
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019558353789525662,
+      "loss": 4.759,
+      "step": 4683
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019554108442862837,
+      "loss": 4.2805,
+      "step": 4684
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019549843254592736,
+      "loss": 4.6495,
+      "step": 4685
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019545558233573134,
+      "loss": 3.4187,
+      "step": 4686
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001954125338870301,
+      "loss": 2.9481,
+      "step": 4687
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019536928728922494,
+      "loss": 4.8499,
+      "step": 4688
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019532584263212878,
+      "loss": 3.3904,
+      "step": 4689
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019528220000596575,
+      "loss": 3.8244,
+      "step": 4690
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019523835950137125,
+      "loss": 4.334,
+      "step": 4691
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001951943212093916,
+      "loss": 5.7356,
+      "step": 4692
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019515008522148376,
+      "loss": 2.8609,
+      "step": 4693
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019510565162951537,
+      "loss": 3.5068,
+      "step": 4694
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019506102052576442,
+      "loss": 4.4729,
+      "step": 4695
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019501619200291908,
+      "loss": 4.3097,
+      "step": 4696
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001949711661540775,
+      "loss": 3.3344,
+      "step": 4697
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001949259430727477,
+      "loss": 4.2035,
+      "step": 4698
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019488052285284715,
+      "loss": 3.113,
+      "step": 4699
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001948349055887029,
+      "loss": 3.4468,
+      "step": 4700
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019478909137505112,
+      "loss": 5.3337,
+      "step": 4701
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019474308030703702,
+      "loss": 4.6425,
+      "step": 4702
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019469687248021463,
+      "loss": 4.1966,
+      "step": 4703
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019465046799054659,
+      "loss": 3.1606,
+      "step": 4704
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019460386693440396,
+      "loss": 3.738,
+      "step": 4705
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019455706940856604,
+      "loss": 3.6957,
+      "step": 4706
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019451007551022015,
+      "loss": 5.5263,
+      "step": 4707
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019446288533696144,
+      "loss": 3.6161,
+      "step": 4708
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019441549898679264,
+      "loss": 4.7888,
+      "step": 4709
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001943679165581239,
+      "loss": 3.83,
+      "step": 4710
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019432013814977262,
+      "loss": 2.1063,
+      "step": 4711
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019427216386096313,
+      "loss": 4.29,
+      "step": 4712
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019422399379132664,
+      "loss": 4.4811,
+      "step": 4713
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019417562804090088,
+      "loss": 2.9428,
+      "step": 4714
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019412706671012998,
+      "loss": 4.8739,
+      "step": 4715
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019407830989986432,
+      "loss": 4.5482,
+      "step": 4716
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001940293577113601,
+      "loss": 5.3828,
+      "step": 4717
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001939802102462794,
+      "loss": 3.8469,
+      "step": 4718
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019393086760668981,
+      "loss": 4.4598,
+      "step": 4719
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019388132989506422,
+      "loss": 2.7338,
+      "step": 4720
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019383159721428065,
+      "loss": 5.9627,
+      "step": 4721
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019378166966762205,
+      "loss": 5.4094,
+      "step": 4722
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000193731547358776,
+      "loss": 4.6663,
+      "step": 4723
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019368123039183467,
+      "loss": 4.6156,
+      "step": 4724
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019363071887129436,
+      "loss": 3.8729,
+      "step": 4725
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019358001290205543,
+      "loss": 2.3321,
+      "step": 4726
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019352911258942217,
+      "loss": 3.9426,
+      "step": 4727
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019347801803910238,
+      "loss": 5.9469,
+      "step": 4728
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001934267293572072,
+      "loss": 4.4062,
+      "step": 4729
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019337524665025116,
+      "loss": 4.7306,
+      "step": 4730
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019332357002515145,
+      "loss": 4.8442,
+      "step": 4731
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019327169958922814,
+      "loss": 5.7048,
+      "step": 4732
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001932196354502038,
+      "loss": 4.7644,
+      "step": 4733
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001931673777162032,
+      "loss": 4.2676,
+      "step": 4734
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019311492649575323,
+      "loss": 4.5383,
+      "step": 4735
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019306228189778257,
+      "loss": 4.0972,
+      "step": 4736
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001930094440316215,
+      "loss": 3.8947,
+      "step": 4737
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001929564130070017,
+      "loss": 2.2727,
+      "step": 4738
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000192903188934056,
+      "loss": 2.9418,
+      "step": 4739
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019284977192331808,
+      "loss": 4.4887,
+      "step": 4740
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019279616208572233,
+      "loss": 2.8125,
+      "step": 4741
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001927423595326037,
+      "loss": 3.7402,
+      "step": 4742
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019268836437569715,
+      "loss": 3.1491,
+      "step": 4743
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019263417672713787,
+      "loss": 4.4623,
+      "step": 4744
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019257979669946062,
+      "loss": 3.2421,
+      "step": 4745
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019252522440559982,
+      "loss": 3.2385,
+      "step": 4746
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019247045995888907,
+      "loss": 6.0402,
+      "step": 4747
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019241550347306113,
+      "loss": 4.0862,
+      "step": 4748
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019236035506224743,
+      "loss": 3.1312,
+      "step": 4749
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019230501484097823,
+      "loss": 2.9868,
+      "step": 4750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019224948292418192,
+      "loss": 4.1637,
+      "step": 4751
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019219375942718507,
+      "loss": 5.1526,
+      "step": 4752
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019213784446571209,
+      "loss": 4.3704,
+      "step": 4753
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019208173815588503,
+      "loss": 4.2088,
+      "step": 4754
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019202544061422332,
+      "loss": 4.5256,
+      "step": 4755
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019196895195764364,
+      "loss": 4.3472,
+      "step": 4756
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019191227230345935,
+      "loss": 4.3702,
+      "step": 4757
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019185540176938065,
+      "loss": 4.5946,
+      "step": 4758
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019179834047351407,
+      "loss": 3.8239,
+      "step": 4759
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019174108853436232,
+      "loss": 3.2113,
+      "step": 4760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001916836460708241,
+      "loss": 2.5576,
+      "step": 4761
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019162601320219365,
+      "loss": 4.1378,
+      "step": 4762
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019156819004816073,
+      "loss": 4.2141,
+      "step": 4763
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019151017672881028,
+      "loss": 5.2917,
+      "step": 4764
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001914519733646222,
+      "loss": 4.3295,
+      "step": 4765
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019139358007647087,
+      "loss": 4.6898,
+      "step": 4766
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001913349969856253,
+      "loss": 3.5584,
+      "step": 4767
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019127622421374866,
+      "loss": 5.5052,
+      "step": 4768
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019121726188289797,
+      "loss": 2.9952,
+      "step": 4769
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019115811011552388,
+      "loss": 4.4134,
+      "step": 4770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019109876903447063,
+      "loss": 2.9504,
+      "step": 4771
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019103923876297545,
+      "loss": 6.1773,
+      "step": 4772
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019097951942466854,
+      "loss": 4.8283,
+      "step": 4773
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019091961114357277,
+      "loss": 2.4917,
+      "step": 4774
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019085951404410333,
+      "loss": 3.7444,
+      "step": 4775
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019079922825106753,
+      "loss": 4.2794,
+      "step": 4776
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001907387538896647,
+      "loss": 4.8856,
+      "step": 4777
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019067809108548562,
+      "loss": 2.9558,
+      "step": 4778
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001906172399645125,
+      "loss": 3.2927,
+      "step": 4779
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019055620065311864,
+      "loss": 4.0681,
+      "step": 4780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019049497327806812,
+      "loss": 3.7556,
+      "step": 4781
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019043355796651563,
+      "loss": 4.4989,
+      "step": 4782
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019037195484600616,
+      "loss": 3.7413,
+      "step": 4783
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019031016404447469,
+      "loss": 5.8595,
+      "step": 4784
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019024818569024602,
+      "loss": 3.3673,
+      "step": 4785
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019018601991203435,
+      "loss": 3.4851,
+      "step": 4786
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019012366683894335,
+      "loss": 4.1583,
+      "step": 4787
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019006112660046535,
+      "loss": 2.6726,
+      "step": 4788
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018999839932648156,
+      "loss": 4.5744,
+      "step": 4789
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018993548514726163,
+      "loss": 4.7824,
+      "step": 4790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018987238419346326,
+      "loss": 4.2526,
+      "step": 4791
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018980909659613215,
+      "loss": 5.1707,
+      "step": 4792
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018974562248670157,
+      "loss": 4.7457,
+      "step": 4793
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018968196199699205,
+      "loss": 3.799,
+      "step": 4794
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001896181152592114,
+      "loss": 4.4822,
+      "step": 4795
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018955408240595398,
+      "loss": 2.4186,
+      "step": 4796
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018948986357020074,
+      "loss": 4.9375,
+      "step": 4797
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000189425458885319,
+      "loss": 5.2528,
+      "step": 4798
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001893608684850619,
+      "loss": 3.1355,
+      "step": 4799
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018929609250356827,
+      "loss": 4.2218,
+      "step": 4800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001892311310753625,
+      "loss": 3.6832,
+      "step": 4801
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018916598433535393,
+      "loss": 3.6776,
+      "step": 4802
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001891006524188368,
+      "loss": 4.0703,
+      "step": 4803
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018903513546149,
+      "loss": 5.6991,
+      "step": 4804
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018896943359937666,
+      "loss": 4.6599,
+      "step": 4805
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001889035469689438,
+      "loss": 5.862,
+      "step": 4806
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018883747570702227,
+      "loss": 3.1315,
+      "step": 4807
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018877121995082642,
+      "loss": 2.8711,
+      "step": 4808
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001887047798379536,
+      "loss": 3.7359,
+      "step": 4809
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018863815550638417,
+      "loss": 3.3968,
+      "step": 4810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018857134709448093,
+      "loss": 2.9151,
+      "step": 4811
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001885043547409891,
+      "loss": 4.5717,
+      "step": 4812
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018843717858503571,
+      "loss": 5.7119,
+      "step": 4813
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018836981876612976,
+      "loss": 6.0555,
+      "step": 4814
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018830227542416153,
+      "loss": 4.4339,
+      "step": 4815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018823454869940242,
+      "loss": 3.902,
+      "step": 4816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018816663873250471,
+      "loss": 5.0169,
+      "step": 4817
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018809854566450126,
+      "loss": 4.005,
+      "step": 4818
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018803026963680513,
+      "loss": 5.8227,
+      "step": 4819
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001879618107912093,
+      "loss": 4.7015,
+      "step": 4820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018789316926988656,
+      "loss": 4.5155,
+      "step": 4821
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000187824345215389,
+      "loss": 3.7961,
+      "step": 4822
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001877553387706477,
+      "loss": 3.7752,
+      "step": 4823
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018768615007897272,
+      "loss": 2.8343,
+      "step": 4824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018761677928405242,
+      "loss": 1.4969,
+      "step": 4825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018754722652995347,
+      "loss": 1.7137,
+      "step": 4826
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018747749196112033,
+      "loss": 4.5466,
+      "step": 4827
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018740757572237508,
+      "loss": 3.7563,
+      "step": 4828
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001873374779589171,
+      "loss": 4.8991,
+      "step": 4829
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018726719881632277,
+      "loss": 3.7446,
+      "step": 4830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018719673844054514,
+      "loss": 1.7569,
+      "step": 4831
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018712609697791354,
+      "loss": 3.6313,
+      "step": 4832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018705527457513353,
+      "loss": 2.9018,
+      "step": 4833
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018698427137928638,
+      "loss": 4.8217,
+      "step": 4834
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018691308753782877,
+      "loss": 3.7468,
+      "step": 4835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001868417231985926,
+      "loss": 4.4987,
+      "step": 4836
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018677017850978467,
+      "loss": 4.3153,
+      "step": 4837
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001866984536199861,
+      "loss": 2.3945,
+      "step": 4838
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001866265486781525,
+      "loss": 4.6462,
+      "step": 4839
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018655446383361333,
+      "loss": 3.6951,
+      "step": 4840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001864821992360716,
+      "loss": 2.773,
+      "step": 4841
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018640975503560366,
+      "loss": 3.847,
+      "step": 4842
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018633713138265884,
+      "loss": 2.4172,
+      "step": 4843
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018626432842805922,
+      "loss": 3.9036,
+      "step": 4844
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018619134632299913,
+      "loss": 5.0436,
+      "step": 4845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018611818521904505,
+      "loss": 4.0446,
+      "step": 4846
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018604484526813523,
+      "loss": 2.3889,
+      "step": 4847
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018597132662257916,
+      "loss": 2.7073,
+      "step": 4848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018589762943505765,
+      "loss": 3.3589,
+      "step": 4849
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018582375385862215,
+      "loss": 5.6184,
+      "step": 4850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018574970004669466,
+      "loss": 3.8559,
+      "step": 4851
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018567546815306726,
+      "loss": 5.0778,
+      "step": 4852
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018560105833190194,
+      "loss": 3.7123,
+      "step": 4853
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001855264707377302,
+      "loss": 2.6153,
+      "step": 4854
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018545170552545265,
+      "loss": 3.0323,
+      "step": 4855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018537676285033884,
+      "loss": 3.4462,
+      "step": 4856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018530164286802693,
+      "loss": 2.2368,
+      "step": 4857
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001852263457345231,
+      "loss": 3.8319,
+      "step": 4858
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001851508716062017,
+      "loss": 2.2509,
+      "step": 4859
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001850752206398044,
+      "loss": 2.6043,
+      "step": 4860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018499939299244025,
+      "loss": 3.5395,
+      "step": 4861
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018492338882158523,
+      "loss": 3.1969,
+      "step": 4862
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00018484720828508195,
+      "loss": 3.8462,
+      "step": 4863
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018477085154113918,
+      "loss": 4.633,
+      "step": 4864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018469431874833166,
+      "loss": 4.3503,
+      "step": 4865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018461761006559983,
+      "loss": 3.3927,
+      "step": 4866
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018454072565224933,
+      "loss": 3.0207,
+      "step": 4867
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001844636656679508,
+      "loss": 4.7213,
+      "step": 4868
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001843864302727394,
+      "loss": 4.0067,
+      "step": 4869
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001843090196270147,
+      "loss": 4.196,
+      "step": 4870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018423143389154013,
+      "loss": 4.1943,
+      "step": 4871
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018415367322744284,
+      "loss": 3.735,
+      "step": 4872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018407573779621314,
+      "loss": 3.7576,
+      "step": 4873
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018399762775970442,
+      "loss": 2.7209,
+      "step": 4874
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018391934328013254,
+      "loss": 3.5828,
+      "step": 4875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018384088452007578,
+      "loss": 4.135,
+      "step": 4876
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018376225164247424,
+      "loss": 4.4329,
+      "step": 4877
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018368344481062976,
+      "loss": 4.3817,
+      "step": 4878
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001836044641882053,
+      "loss": 3.9145,
+      "step": 4879
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018352530993922481,
+      "loss": 3.0001,
+      "step": 4880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001834459822280728,
+      "loss": 3.6218,
+      "step": 4881
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018336648121949407,
+      "loss": 4.6139,
+      "step": 4882
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001832868070785932,
+      "loss": 2.4778,
+      "step": 4883
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018320695997083443,
+      "loss": 3.8177,
+      "step": 4884
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001831269400620412,
+      "loss": 4.122,
+      "step": 4885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001830467475183958,
+      "loss": 3.758,
+      "step": 4886
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018296638250643902,
+      "loss": 1.9357,
+      "step": 4887
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018288584519306986,
+      "loss": 3.4861,
+      "step": 4888
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018280513574554514,
+      "loss": 4.4758,
+      "step": 4889
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018272425433147913,
+      "loss": 5.0186,
+      "step": 4890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018264320111884333,
+      "loss": 6.1832,
+      "step": 4891
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018256197627596586,
+      "loss": 4.2786,
+      "step": 4892
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018248057997153139,
+      "loss": 5.0641,
+      "step": 4893
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001823990123745807,
+      "loss": 2.2892,
+      "step": 4894
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001823172736545102,
+      "loss": 3.0156,
+      "step": 4895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018223536398107178,
+      "loss": 3.9039,
+      "step": 4896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001821532835243723,
+      "loss": 5.4426,
+      "step": 4897
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018207103245487336,
+      "loss": 5.1285,
+      "step": 4898
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001819886109433908,
+      "loss": 2.4981,
+      "step": 4899
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001819060191610945,
+      "loss": 5.8613,
+      "step": 4900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018182325727950785,
+      "loss": 3.7446,
+      "step": 4901
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001817403254705076,
+      "loss": 1.8529,
+      "step": 4902
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018165722390632333,
+      "loss": 4.4648,
+      "step": 4903
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018157395275953726,
+      "loss": 3.1146,
+      "step": 4904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018149051220308366,
+      "loss": 4.5594,
+      "step": 4905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018140690241024874,
+      "loss": 3.0786,
+      "step": 4906
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001813231235546701,
+      "loss": 2.7446,
+      "step": 4907
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018123917581033645,
+      "loss": 4.7648,
+      "step": 4908
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001811550593515872,
+      "loss": 4.3678,
+      "step": 4909
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018107077435311237,
+      "loss": 3.1629,
+      "step": 4910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001809863209899517,
+      "loss": 3.917,
+      "step": 4911
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001809016994374947,
+      "loss": 5.7011,
+      "step": 4912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018081690987148025,
+      "loss": 4.5805,
+      "step": 4913
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018073195246799605,
+      "loss": 3.1438,
+      "step": 4914
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018064682740347835,
+      "loss": 3.6462,
+      "step": 4915
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018056153485471166,
+      "loss": 5.2719,
+      "step": 4916
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018047607499882827,
+      "loss": 2.9321,
+      "step": 4917
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018039044801330793,
+      "loss": 4.0793,
+      "step": 4918
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018030465407597754,
+      "loss": 5.5309,
+      "step": 4919
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018021869336501064,
+      "loss": 4.2869,
+      "step": 4920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018013256605892714,
+      "loss": 3.9117,
+      "step": 4921
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00018004627233659303,
+      "loss": 6.7701,
+      "step": 4922
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017995981237721966,
+      "loss": 3.3593,
+      "step": 4923
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017987318636036384,
+      "loss": 5.019,
+      "step": 4924
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001797863944659272,
+      "loss": 3.3982,
+      "step": 4925
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001796994368741558,
+      "loss": 4.2332,
+      "step": 4926
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017961231376563985,
+      "loss": 3.5086,
+      "step": 4927
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017952502532131326,
+      "loss": 3.4021,
+      "step": 4928
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001794375717224534,
+      "loss": 5.59,
+      "step": 4929
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017934995315068051,
+      "loss": 5.1558,
+      "step": 4930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017926216978795758,
+      "loss": 3.7569,
+      "step": 4931
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001791742218165896,
+      "loss": 4.5683,
+      "step": 4932
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017908610941922364,
+      "loss": 5.0127,
+      "step": 4933
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017899783277884815,
+      "loss": 3.9292,
+      "step": 4934
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017890939207879268,
+      "loss": 5.3212,
+      "step": 4935
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017882078750272747,
+      "loss": 3.4121,
+      "step": 4936
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017873201923466317,
+      "loss": 4.4867,
+      "step": 4937
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001786430874589503,
+      "loss": 3.7071,
+      "step": 4938
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017855399236027894,
+      "loss": 3.9198,
+      "step": 4939
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017846473412367852,
+      "loss": 4.3215,
+      "step": 4940
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017837531293451683,
+      "loss": 4.1698,
+      "step": 4941
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001782857289785007,
+      "loss": 4.7025,
+      "step": 4942
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017819598244167456,
+      "loss": 5.6089,
+      "step": 4943
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017810607351042058,
+      "loss": 3.1882,
+      "step": 4944
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017801600237145823,
+      "loss": 1.7,
+      "step": 4945
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017792576921184372,
+      "loss": 3.8326,
+      "step": 4946
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017783537421896988,
+      "loss": 3.7483,
+      "step": 4947
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017774481758056553,
+      "loss": 2.29,
+      "step": 4948
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017765409948469532,
+      "loss": 4.5148,
+      "step": 4949
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017756322011975906,
+      "loss": 4.5769,
+      "step": 4950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017747217967449156,
+      "loss": 5.2873,
+      "step": 4951
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017738097833796217,
+      "loss": 4.6866,
+      "step": 4952
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017728961629957427,
+      "loss": 3.8294,
+      "step": 4953
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017719809374906514,
+      "loss": 2.25,
+      "step": 4954
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017710641087650524,
+      "loss": 3.8263,
+      "step": 4955
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001770145678722981,
+      "loss": 5.0888,
+      "step": 4956
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001769225649271797,
+      "loss": 3.2542,
+      "step": 4957
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001768304022322183,
+      "loss": 2.4807,
+      "step": 4958
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017673807997881382,
+      "loss": 3.8847,
+      "step": 4959
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017664559835869765,
+      "loss": 4.3962,
+      "step": 4960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017655295756393202,
+      "loss": 5.4666,
+      "step": 4961
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017646015778690984,
+      "loss": 3.7963,
+      "step": 4962
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017636719922035416,
+      "loss": 3.1679,
+      "step": 4963
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017627408205731765,
+      "loss": 4.2632,
+      "step": 4964
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001761808064911826,
+      "loss": 3.2951,
+      "step": 4965
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001760873727156601,
+      "loss": 4.7365,
+      "step": 4966
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017599378092478983,
+      "loss": 4.1178,
+      "step": 4967
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001759000313129397,
+      "loss": 2.9431,
+      "step": 4968
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017580612407480526,
+      "loss": 1.6536,
+      "step": 4969
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017571205940540948,
+      "loss": 4.5546,
+      "step": 4970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017561783750010232,
+      "loss": 1.9848,
+      "step": 4971
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017552345855456026,
+      "loss": 3.5626,
+      "step": 4972
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017542892276478563,
+      "loss": 4.3141,
+      "step": 4973
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017533423032710702,
+      "loss": 2.2063,
+      "step": 4974
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017523938143817805,
+      "loss": 3.2777,
+      "step": 4975
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017514437629497715,
+      "loss": 4.512,
+      "step": 4976
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001750492150948075,
+      "loss": 2.8381,
+      "step": 4977
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017495389803529613,
+      "loss": 3.9529,
+      "step": 4978
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001748584253143939,
+      "loss": 3.4692,
+      "step": 4979
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001747627971303749,
+      "loss": 3.4197,
+      "step": 4980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017466701368183611,
+      "loss": 3.3755,
+      "step": 4981
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017457107516769693,
+      "loss": 2.9952,
+      "step": 4982
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017447498178719878,
+      "loss": 4.2556,
+      "step": 4983
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017437873373990473,
+      "loss": 2.681,
+      "step": 4984
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017428233122569907,
+      "loss": 4.2952,
+      "step": 4985
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017418577444478693,
+      "loss": 3.8992,
+      "step": 4986
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017408906359769354,
+      "loss": 2.6421,
+      "step": 4987
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017399219888526441,
+      "loss": 3.8486,
+      "step": 4988
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017389518050866447,
+      "loss": 4.5627,
+      "step": 4989
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001737980086693778,
+      "loss": 3.4299,
+      "step": 4990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001737006835692071,
+      "loss": 2.4498,
+      "step": 4991
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017360320541027344,
+      "loss": 3.4587,
+      "step": 4992
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017350557439501575,
+      "loss": 2.6718,
+      "step": 4993
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001734077907261905,
+      "loss": 3.5918,
+      "step": 4994
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017330985460687086,
+      "loss": 4.234,
+      "step": 4995
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017321176624044693,
+      "loss": 4.1844,
+      "step": 4996
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017311352583062489,
+      "loss": 3.8238,
+      "step": 4997
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017301513358142665,
+      "loss": 2.8923,
+      "step": 4998
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017291658969718947,
+      "loss": 5.5728,
+      "step": 4999
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017281789438256554,
+      "loss": 2.3788,
+      "step": 5000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017271904784252147,
+      "loss": 3.2955,
+      "step": 5001
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017262005028233806,
+      "loss": 3.8309,
+      "step": 5002
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017252090190760964,
+      "loss": 3.6911,
+      "step": 5003
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001724216029242437,
+      "loss": 3.6635,
+      "step": 5004
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001723221535384605,
+      "loss": 3.5878,
+      "step": 5005
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017222255395679293,
+      "loss": 4.7647,
+      "step": 5006
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001721228043860855,
+      "loss": 4.5666,
+      "step": 5007
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017202290503349436,
+      "loss": 4.6686,
+      "step": 5008
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001719228561064866,
+      "loss": 3.0453,
+      "step": 5009
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017182265781284,
+      "loss": 5.4851,
+      "step": 5010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017172231036064253,
+      "loss": 2.0527,
+      "step": 5011
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017162181395829205,
+      "loss": 3.5954,
+      "step": 5012
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017152116881449553,
+      "loss": 4.0698,
+      "step": 5013
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017142037513826904,
+      "loss": 3.8109,
+      "step": 5014
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017131943313893698,
+      "loss": 3.8748,
+      "step": 5015
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017121834302613186,
+      "loss": 4.0074,
+      "step": 5016
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001711171050097938,
+      "loss": 5.1391,
+      "step": 5017
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017101571930017,
+      "loss": 2.304,
+      "step": 5018
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017091418610781442,
+      "loss": 1.7233,
+      "step": 5019
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017081250564358728,
+      "loss": 1.4411,
+      "step": 5020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017071067811865479,
+      "loss": 4.6678,
+      "step": 5021
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017060870374448837,
+      "loss": 4.8272,
+      "step": 5022
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017050658273286452,
+      "loss": 3.4983,
+      "step": 5023
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001704043152958643,
+      "loss": 5.4258,
+      "step": 5024
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017030190164587277,
+      "loss": 1.4979,
+      "step": 5025
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017019934199557875,
+      "loss": 3.8288,
+      "step": 5026
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001700966365579741,
+      "loss": 4.3173,
+      "step": 5027
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016999378554635353,
+      "loss": 2.4822,
+      "step": 5028
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016989078917431424,
+      "loss": 3.4114,
+      "step": 5029
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016978764765575503,
+      "loss": 3.9887,
+      "step": 5030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016968436120487634,
+      "loss": 4.1686,
+      "step": 5031
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016958093003617948,
+      "loss": 4.7902,
+      "step": 5032
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016947735436446636,
+      "loss": 5.0641,
+      "step": 5033
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016937363440483897,
+      "loss": 3.9858,
+      "step": 5034
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016926977037269903,
+      "loss": 3.3211,
+      "step": 5035
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016916576248374727,
+      "loss": 2.5066,
+      "step": 5036
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016906161095398337,
+      "loss": 4.6128,
+      "step": 5037
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001689573159997051,
+      "loss": 3.1221,
+      "step": 5038
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016885287783750848,
+      "loss": 3.1078,
+      "step": 5039
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016874829668428665,
+      "loss": 4.3869,
+      "step": 5040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016864357275722967,
+      "loss": 3.0741,
+      "step": 5041
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016853870627382412,
+      "loss": 1.7854,
+      "step": 5042
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016843369745185287,
+      "loss": 4.3387,
+      "step": 5043
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001683285465093941,
+      "loss": 3.6764,
+      "step": 5044
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00016822325366482133,
+      "loss": 2.7899,
+      "step": 5045
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001681178191368027,
+      "loss": 2.8889,
+      "step": 5046
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001680122431443006,
+      "loss": 2.2785,
+      "step": 5047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016790652590657122,
+      "loss": 3.349,
+      "step": 5048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016780066764316418,
+      "loss": 4.1046,
+      "step": 5049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001676946685739217,
+      "loss": 2.4913,
+      "step": 5050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001675885289189787,
+      "loss": 4.0009,
+      "step": 5051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001674822488987619,
+      "loss": 4.7362,
+      "step": 5052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016737582873398963,
+      "loss": 2.0399,
+      "step": 5053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001672692686456712,
+      "loss": 1.8094,
+      "step": 5054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016716256885510657,
+      "loss": 3.3438,
+      "step": 5055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016705572958388576,
+      "loss": 2.7944,
+      "step": 5056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016694875105388853,
+      "loss": 5.8219,
+      "step": 5057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016684163348728366,
+      "loss": 5.3545,
+      "step": 5058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016673437710652898,
+      "loss": 3.1315,
+      "step": 5059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016662698213437032,
+      "loss": 4.4326,
+      "step": 5060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016651944879384154,
+      "loss": 3.8074,
+      "step": 5061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016641177730826372,
+      "loss": 2.9782,
+      "step": 5062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016630396790124493,
+      "loss": 3.8264,
+      "step": 5063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016619602079667958,
+      "loss": 4.8711,
+      "step": 5064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016608793621874815,
+      "loss": 4.1909,
+      "step": 5065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016597971439191657,
+      "loss": 4.7992,
+      "step": 5066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001658713555409357,
+      "loss": 5.3857,
+      "step": 5067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016576285989084108,
+      "loss": 2.3987,
+      "step": 5068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016565422766695237,
+      "loss": 2.1264,
+      "step": 5069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001655454590948727,
+      "loss": 5.1823,
+      "step": 5070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016543655440048877,
+      "loss": 3.0757,
+      "step": 5071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016532751380996948,
+      "loss": 4.2348,
+      "step": 5072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016521833754976615,
+      "loss": 3.7243,
+      "step": 5073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016510902584661191,
+      "loss": 3.7353,
+      "step": 5074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016499957892752117,
+      "loss": 0.8298,
+      "step": 5075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016488999701978903,
+      "loss": 3.3258,
+      "step": 5076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016478028035099104,
+      "loss": 4.9958,
+      "step": 5077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016467042914898258,
+      "loss": 4.8752,
+      "step": 5078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016456044364189846,
+      "loss": 3.6574,
+      "step": 5079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016445032405815242,
+      "loss": 3.4756,
+      "step": 5080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016434007062643646,
+      "loss": 4.2212,
+      "step": 5081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016422968357572074,
+      "loss": 3.3647,
+      "step": 5082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016411916313525295,
+      "loss": 5.2634,
+      "step": 5083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016400850953455763,
+      "loss": 4.1541,
+      "step": 5084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000163897723003436,
+      "loss": 4.169,
+      "step": 5085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016378680377196524,
+      "loss": 2.5202,
+      "step": 5086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016367575207049822,
+      "loss": 3.3783,
+      "step": 5087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016356456812966288,
+      "loss": 2.4066,
+      "step": 5088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016345325218036183,
+      "loss": 4.4972,
+      "step": 5089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016334180445377162,
+      "loss": 4.0326,
+      "step": 5090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001632302251813427,
+      "loss": 4.4591,
+      "step": 5091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016311851459479865,
+      "loss": 3.2998,
+      "step": 5092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016300667292613578,
+      "loss": 4.744,
+      "step": 5093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016289470040762254,
+      "loss": 4.1044,
+      "step": 5094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001627825972717992,
+      "loss": 4.0566,
+      "step": 5095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016267036375147728,
+      "loss": 4.0456,
+      "step": 5096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016255800007973904,
+      "loss": 3.1397,
+      "step": 5097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016244550648993717,
+      "loss": 3.248,
+      "step": 5098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016233288321569386,
+      "loss": 1.8978,
+      "step": 5099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016222013049090094,
+      "loss": 2.7352,
+      "step": 5100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016210724854971896,
+      "loss": 3.2317,
+      "step": 5101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001619942376265766,
+      "loss": 4.7633,
+      "step": 5102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001618810979561711,
+      "loss": 4.5529,
+      "step": 5103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016176782977346624,
+      "loss": 2.4943,
+      "step": 5104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016165443331369318,
+      "loss": 2.2056,
+      "step": 5105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016154090881234927,
+      "loss": 2.4311,
+      "step": 5106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016142725650519793,
+      "loss": 4.2831,
+      "step": 5107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001613134766282679,
+      "loss": 3.8372,
+      "step": 5108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016119956941785288,
+      "loss": 4.0459,
+      "step": 5109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016108553511051104,
+      "loss": 3.5334,
+      "step": 5110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001609713739430644,
+      "loss": 4.2361,
+      "step": 5111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016085708615259865,
+      "loss": 3.4836,
+      "step": 5112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016074267197646215,
+      "loss": 3.735,
+      "step": 5113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016062813165226595,
+      "loss": 4.0138,
+      "step": 5114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016051346541788298,
+      "loss": 5.4345,
+      "step": 5115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016039867351144778,
+      "loss": 3.597,
+      "step": 5116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016028375617135573,
+      "loss": 4.0268,
+      "step": 5117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016016871363626277,
+      "loss": 3.8209,
+      "step": 5118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016005354614508487,
+      "loss": 4.6137,
+      "step": 5119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015993825393699745,
+      "loss": 5.5547,
+      "step": 5120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015982283725143504,
+      "loss": 3.9083,
+      "step": 5121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015970729632809046,
+      "loss": 3.1732,
+      "step": 5122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001595916314069148,
+      "loss": 3.7524,
+      "step": 5123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015947584272811644,
+      "loss": 4.273,
+      "step": 5124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015935993053216096,
+      "loss": 4.0869,
+      "step": 5125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001592438950597704,
+      "loss": 3.7319,
+      "step": 5126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015912773655192273,
+      "loss": 3.2412,
+      "step": 5127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015901145524985154,
+      "loss": 3.2314,
+      "step": 5128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015889505139504546,
+      "loss": 4.8589,
+      "step": 5129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001587785252292474,
+      "loss": 2.9181,
+      "step": 5130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015866187699445456,
+      "loss": 3.7524,
+      "step": 5131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015854510693291753,
+      "loss": 0.7617,
+      "step": 5132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015842821528713988,
+      "loss": 4.6331,
+      "step": 5133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015831120229987762,
+      "loss": 4.2515,
+      "step": 5134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015819406821413907,
+      "loss": 4.2855,
+      "step": 5135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001580768132731837,
+      "loss": 3.5358,
+      "step": 5136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001579594377205221,
+      "loss": 2.4391,
+      "step": 5137
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015784194179991533,
+      "loss": 4.3353,
+      "step": 5138
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015772432575537443,
+      "loss": 3.2982,
+      "step": 5139
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015760658983115993,
+      "loss": 4.9506,
+      "step": 5140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015748873427178124,
+      "loss": 4.5101,
+      "step": 5141
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015737075932199633,
+      "loss": 3.3867,
+      "step": 5142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015725266522681111,
+      "loss": 4.0275,
+      "step": 5143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001571344522314788,
+      "loss": 3.4901,
+      "step": 5144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015701612058149962,
+      "loss": 4.2036,
+      "step": 5145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015689767052262026,
+      "loss": 3.0674,
+      "step": 5146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001567791023008333,
+      "loss": 3.8144,
+      "step": 5147
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001566604161623766,
+      "loss": 4.0002,
+      "step": 5148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001565416123537331,
+      "loss": 5.4438,
+      "step": 5149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015642269112162995,
+      "loss": 4.0084,
+      "step": 5150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015630365271303827,
+      "loss": 3.3695,
+      "step": 5151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015618449737517248,
+      "loss": 4.5043,
+      "step": 5152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015606522535548974,
+      "loss": 4.0694,
+      "step": 5153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015594583690168968,
+      "loss": 5.02,
+      "step": 5154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015582633226171372,
+      "loss": 1.7471,
+      "step": 5155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001557067116837444,
+      "loss": 5.2172,
+      "step": 5156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015558697541620534,
+      "loss": 3.3591,
+      "step": 5157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001554671237077601,
+      "loss": 2.1653,
+      "step": 5158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015534715680731219,
+      "loss": 3.9476,
+      "step": 5159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001552270749640043,
+      "loss": 3.6613,
+      "step": 5160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015510687842721786,
+      "loss": 3.5495,
+      "step": 5161
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015498656744657228,
+      "loss": 5.0775,
+      "step": 5162
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001548661422719249,
+      "loss": 5.0208,
+      "step": 5163
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015474560315337015,
+      "loss": 4.1895,
+      "step": 5164
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015462495034123901,
+      "loss": 3.7124,
+      "step": 5165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001545041840860986,
+      "loss": 4.0916,
+      "step": 5166
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015438330463875177,
+      "loss": 3.458,
+      "step": 5167
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015426231225023628,
+      "loss": 1.5953,
+      "step": 5168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015414120717182454,
+      "loss": 3.0202,
+      "step": 5169
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015401998965502287,
+      "loss": 2.9522,
+      "step": 5170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015389865995157128,
+      "loss": 4.1298,
+      "step": 5171
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015377721831344253,
+      "loss": 3.1721,
+      "step": 5172
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015365566499284206,
+      "loss": 3.898,
+      "step": 5173
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015353400024220717,
+      "loss": 4.372,
+      "step": 5174
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015341222431420657,
+      "loss": 6.8456,
+      "step": 5175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015329033746173978,
+      "loss": 5.1409,
+      "step": 5176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001531683399379368,
+      "loss": 4.9482,
+      "step": 5177
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015304623199615736,
+      "loss": 3.3692,
+      "step": 5178
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001529240138899907,
+      "loss": 1.6799,
+      "step": 5179
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001528016858732546,
+      "loss": 4.4882,
+      "step": 5180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015267924819999527,
+      "loss": 3.9589,
+      "step": 5181
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015255670112448657,
+      "loss": 4.5843,
+      "step": 5182
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015243404490122958,
+      "loss": 3.8921,
+      "step": 5183
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015231127978495215,
+      "loss": 3.4852,
+      "step": 5184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015218840603060797,
+      "loss": 6.0666,
+      "step": 5185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015206542389337675,
+      "loss": 2.7848,
+      "step": 5186
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015194233362866296,
+      "loss": 2.0868,
+      "step": 5187
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015181913549209585,
+      "loss": 5.8848,
+      "step": 5188
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015169582973952855,
+      "loss": 3.4499,
+      "step": 5189
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001515724166270378,
+      "loss": 5.032,
+      "step": 5190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015144889641092316,
+      "loss": 6.1475,
+      "step": 5191
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001513252693477068,
+      "loss": 5.4003,
+      "step": 5192
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015120153569413253,
+      "loss": 4.9297,
+      "step": 5193
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015107769570716574,
+      "loss": 4.3434,
+      "step": 5194
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001509537496439926,
+      "loss": 4.3073,
+      "step": 5195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015082969776201955,
+      "loss": 4.9671,
+      "step": 5196
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015070554031887282,
+      "loss": 1.2956,
+      "step": 5197
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015058127757239768,
+      "loss": 2.5422,
+      "step": 5198
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001504569097806585,
+      "loss": 4.9843,
+      "step": 5199
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015033243720193745,
+      "loss": 2.7615,
+      "step": 5200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015020786009473441,
+      "loss": 1.8259,
+      "step": 5201
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00015008317871776638,
+      "loss": 4.408,
+      "step": 5202
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014995839332996685,
+      "loss": 3.6854,
+      "step": 5203
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014983350419048538,
+      "loss": 3.3402,
+      "step": 5204
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014970851155868693,
+      "loss": 1.3379,
+      "step": 5205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014958341569415153,
+      "loss": 4.3433,
+      "step": 5206
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001494582168566733,
+      "loss": 4.1656,
+      "step": 5207
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014933291530626049,
+      "loss": 3.7862,
+      "step": 5208
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014920751130313454,
+      "loss": 3.1225,
+      "step": 5209
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014908200510772978,
+      "loss": 3.7756,
+      "step": 5210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001489563969806926,
+      "loss": 2.0903,
+      "step": 5211
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001488306871828812,
+      "loss": 3.9365,
+      "step": 5212
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014870487597536485,
+      "loss": 1.7198,
+      "step": 5213
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014857896361942347,
+      "loss": 4.2088,
+      "step": 5214
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014845295037654713,
+      "loss": 4.5903,
+      "step": 5215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014832683650843514,
+      "loss": 2.342,
+      "step": 5216
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014820062227699604,
+      "loss": 1.8545,
+      "step": 5217
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014807430794434675,
+      "loss": 2.554,
+      "step": 5218
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014794789377281197,
+      "loss": 4.4736,
+      "step": 5219
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001478213800249239,
+      "loss": 4.6652,
+      "step": 5220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014769476696342143,
+      "loss": 3.1749,
+      "step": 5221
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001475680548512497,
+      "loss": 4.3871,
+      "step": 5222
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014744124395155964,
+      "loss": 2.8015,
+      "step": 5223
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001473143345277073,
+      "loss": 5.1991,
+      "step": 5224
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014718732684325324,
+      "loss": 2.8052,
+      "step": 5225
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014706022116196213,
+      "loss": 3.1796,
+      "step": 5226
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014693301774780233,
+      "loss": 4.6011,
+      "step": 5227
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00014680571686494492,
+      "loss": 3.8615,
+      "step": 5228
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001466783187777635,
+      "loss": 3.2001,
+      "step": 5229
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001465508237508336,
+      "loss": 2.4383,
+      "step": 5230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000146423232048932,
+      "loss": 3.3617,
+      "step": 5231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014629554393703632,
+      "loss": 3.8993,
+      "step": 5232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014616775968032431,
+      "loss": 4.1241,
+      "step": 5233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014603987954417346,
+      "loss": 3.641,
+      "step": 5234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001459119037941603,
+      "loss": 4.6794,
+      "step": 5235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014578383269606,
+      "loss": 2.6371,
+      "step": 5236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014565566651584573,
+      "loss": 2.9965,
+      "step": 5237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014552740551968815,
+      "loss": 3.8791,
+      "step": 5238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001453990499739547,
+      "loss": 4.5254,
+      "step": 5239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014527060014520932,
+      "loss": 4.7113,
+      "step": 5240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001451420563002117,
+      "loss": 3.0592,
+      "step": 5241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014501341870591682,
+      "loss": 4.0038,
+      "step": 5242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014488468762947427,
+      "loss": 3.9557,
+      "step": 5243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014475586333822786,
+      "loss": 2.3463,
+      "step": 5244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014462694609971495,
+      "loss": 4.7302,
+      "step": 5245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001444979361816659,
+      "loss": 5.1665,
+      "step": 5246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014436883385200372,
+      "loss": 3.1025,
+      "step": 5247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000144239639378843,
+      "loss": 1.9818,
+      "step": 5248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014411035303048993,
+      "loss": 3.8452,
+      "step": 5249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014398097507544155,
+      "loss": 4.6974,
+      "step": 5250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014385150578238496,
+      "loss": 3.5303,
+      "step": 5251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014372194542019707,
+      "loss": 3.0451,
+      "step": 5252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014359229425794392,
+      "loss": 3.0537,
+      "step": 5253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014346255256488004,
+      "loss": 2.6634,
+      "step": 5254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014333272061044805,
+      "loss": 3.8111,
+      "step": 5255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014320279866427807,
+      "loss": 4.0672,
+      "step": 5256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014307278699618686,
+      "loss": 2.9196,
+      "step": 5257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014294268587617775,
+      "loss": 3.2893,
+      "step": 5258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014281249557443983,
+      "loss": 5.039,
+      "step": 5259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001426822163613473,
+      "loss": 4.4733,
+      "step": 5260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014255184850745908,
+      "loss": 3.4739,
+      "step": 5261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014242139228351798,
+      "loss": 2.8791,
+      "step": 5262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001422908479604508,
+      "loss": 3.6436,
+      "step": 5263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014216021580936678,
+      "loss": 3.8203,
+      "step": 5264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014202949610155787,
+      "loss": 4.0197,
+      "step": 5265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014189868910849772,
+      "loss": 2.7826,
+      "step": 5266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014176779510184133,
+      "loss": 4.9755,
+      "step": 5267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014163681435342435,
+      "loss": 5.0802,
+      "step": 5268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014150574713526267,
+      "loss": 4.1725,
+      "step": 5269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014137459371955154,
+      "loss": 2.5701,
+      "step": 5270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001412433543786654,
+      "loss": 4.2637,
+      "step": 5271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014111202938515716,
+      "loss": 3.795,
+      "step": 5272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014098061901175748,
+      "loss": 3.9021,
+      "step": 5273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014084912353137444,
+      "loss": 4.17,
+      "step": 5274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001407175432170928,
+      "loss": 4.6775,
+      "step": 5275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014058587834217355,
+      "loss": 2.1921,
+      "step": 5276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001404541291800533,
+      "loss": 2.0772,
+      "step": 5277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001403222960043437,
+      "loss": 4.7103,
+      "step": 5278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014019037908883072,
+      "loss": 3.0404,
+      "step": 5279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00014005837870747446,
+      "loss": 3.0448,
+      "step": 5280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013992629513440822,
+      "loss": 5.0132,
+      "step": 5281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013979412864393814,
+      "loss": 2.8968,
+      "step": 5282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013966187951054256,
+      "loss": 5.8303,
+      "step": 5283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001395295480088714,
+      "loss": 4.0228,
+      "step": 5284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013939713441374568,
+      "loss": 3.802,
+      "step": 5285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001392646390001569,
+      "loss": 4.2047,
+      "step": 5286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013913206204326653,
+      "loss": 3.0914,
+      "step": 5287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013899940381840518,
+      "loss": 4.7121,
+      "step": 5288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013886666460107244,
+      "loss": 4.4409,
+      "step": 5289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013873384466693607,
+      "loss": 3.8856,
+      "step": 5290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013860094429183142,
+      "loss": 2.2953,
+      "step": 5291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013846796375176093,
+      "loss": 4.4212,
+      "step": 5292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013833490332289348,
+      "loss": 5.0274,
+      "step": 5293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001382017632815638,
+      "loss": 1.7398,
+      "step": 5294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001380685439042722,
+      "loss": 3.3844,
+      "step": 5295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013793524546768356,
+      "loss": 3.9738,
+      "step": 5296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013780186824862692,
+      "loss": 2.5446,
+      "step": 5297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013766841252409499,
+      "loss": 5.507,
+      "step": 5298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013753487857124354,
+      "loss": 5.5762,
+      "step": 5299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013740126666739083,
+      "loss": 4.1862,
+      "step": 5300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001372675770900169,
+      "loss": 3.9975,
+      "step": 5301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013713381011676317,
+      "loss": 3.5287,
+      "step": 5302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013699996602543172,
+      "loss": 3.4568,
+      "step": 5303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001368660450939849,
+      "loss": 3.8812,
+      "step": 5304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013673204760054458,
+      "loss": 4.2874,
+      "step": 5305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001365979738233916,
+      "loss": 4.0555,
+      "step": 5306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013646382404096532,
+      "loss": 0.9359,
+      "step": 5307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013632959853186283,
+      "loss": 3.2693,
+      "step": 5308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013619529757483853,
+      "loss": 3.7332,
+      "step": 5309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013606092144880365,
+      "loss": 4.1635,
+      "step": 5310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013592647043282518,
+      "loss": 3.3979,
+      "step": 5311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013579194480612596,
+      "loss": 5.8891,
+      "step": 5312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013565734484808365,
+      "loss": 4.9436,
+      "step": 5313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013552267083823032,
+      "loss": 3.9701,
+      "step": 5314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001353879230562518,
+      "loss": 3.7223,
+      "step": 5315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001352531017819871,
+      "loss": 2.6126,
+      "step": 5316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013511820729542796,
+      "loss": 2.5935,
+      "step": 5317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013498323987671805,
+      "loss": 6.0543,
+      "step": 5318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001348481998061526,
+      "loss": 5.9652,
+      "step": 5319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001347130873641776,
+      "loss": 3.2998,
+      "step": 5320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013457790283138936,
+      "loss": 3.9657,
+      "step": 5321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013444264648853404,
+      "loss": 3.379,
+      "step": 5322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013430731861650684,
+      "loss": 6.0568,
+      "step": 5323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013417191949635147,
+      "loss": 3.4326,
+      "step": 5324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013403644940925965,
+      "loss": 4.8349,
+      "step": 5325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013390090863657042,
+      "loss": 4.9274,
+      "step": 5326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001337652974597699,
+      "loss": 3.1091,
+      "step": 5327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013362961616049003,
+      "loss": 3.1538,
+      "step": 5328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013349386502050853,
+      "loss": 3.5052,
+      "step": 5329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013335804432174826,
+      "loss": 4.6312,
+      "step": 5330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001332221543462764,
+      "loss": 4.7674,
+      "step": 5331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013308619537630417,
+      "loss": 3.6905,
+      "step": 5332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001329501676941858,
+      "loss": 3.6136,
+      "step": 5333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013281407158241844,
+      "loss": 3.9131,
+      "step": 5334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013267790732364129,
+      "loss": 5.5158,
+      "step": 5335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001325416752006351,
+      "loss": 1.6049,
+      "step": 5336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013240537549632153,
+      "loss": 4.6163,
+      "step": 5337
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001322690084937626,
+      "loss": 4.9331,
+      "step": 5338
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001321325744761601,
+      "loss": 2.7607,
+      "step": 5339
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013199607372685494,
+      "loss": 3.2111,
+      "step": 5340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001318595065293268,
+      "loss": 5.6551,
+      "step": 5341
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000131722873167193,
+      "loss": 4.9112,
+      "step": 5342
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001315861739242086,
+      "loss": 5.145,
+      "step": 5343
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013144940908426536,
+      "loss": 3.4764,
+      "step": 5344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013131257893139124,
+      "loss": 4.2173,
+      "step": 5345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001311756837497499,
+      "loss": 4.1044,
+      "step": 5346
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013103872382364004,
+      "loss": 3.0985,
+      "step": 5347
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013090169943749476,
+      "loss": 4.37,
+      "step": 5348
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001307646108758811,
+      "loss": 2.1067,
+      "step": 5349
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013062745842349943,
+      "loss": 1.3851,
+      "step": 5350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013049024236518254,
+      "loss": 4.7087,
+      "step": 5351
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013035296298589556,
+      "loss": 2.3288,
+      "step": 5352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00013021562057073508,
+      "loss": 4.8647,
+      "step": 5353
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001300782154049286,
+      "loss": 1.9645,
+      "step": 5354
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012994074777383387,
+      "loss": 4.7826,
+      "step": 5355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012980321796293845,
+      "loss": 3.4407,
+      "step": 5356
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012966562625785897,
+      "loss": 3.5459,
+      "step": 5357
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012952797294434057,
+      "loss": 4.1736,
+      "step": 5358
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012939025830825652,
+      "loss": 3.6079,
+      "step": 5359
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012925248263560726,
+      "loss": 4.3054,
+      "step": 5360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012911464621252004,
+      "loss": 2.7978,
+      "step": 5361
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012897674932524827,
+      "loss": 3.3439,
+      "step": 5362
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012883879226017092,
+      "loss": 5.32,
+      "step": 5363
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012870077530379207,
+      "loss": 4.7931,
+      "step": 5364
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012856269874273986,
+      "loss": 5.9247,
+      "step": 5365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001284245628637665,
+      "loss": 2.9644,
+      "step": 5366
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001282863679537473,
+      "loss": 3.008,
+      "step": 5367
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001281481142996802,
+      "loss": 5.1028,
+      "step": 5368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001280098021886851,
+      "loss": 1.9025,
+      "step": 5369
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012787143190800325,
+      "loss": 3.626,
+      "step": 5370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012773300374499687,
+      "loss": 2.5222,
+      "step": 5371
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012759451798714815,
+      "loss": 3.1501,
+      "step": 5372
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012745597492205918,
+      "loss": 1.0303,
+      "step": 5373
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001273173748374507,
+      "loss": 4.5042,
+      "step": 5374
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001271787180211622,
+      "loss": 4.9172,
+      "step": 5375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012704000476115084,
+      "loss": 3.0602,
+      "step": 5376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000126901235345491,
+      "loss": 4.5227,
+      "step": 5377
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001267624100623737,
+      "loss": 3.4463,
+      "step": 5378
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012662352920010593,
+      "loss": 4.7396,
+      "step": 5379
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012648459304711028,
+      "loss": 5.5229,
+      "step": 5380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012634560189192398,
+      "loss": 2.8433,
+      "step": 5381
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012620655602319858,
+      "loss": 4.2725,
+      "step": 5382
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012606745572969915,
+      "loss": 2.8557,
+      "step": 5383
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012592830130030388,
+      "loss": 5.1172,
+      "step": 5384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012578909302400342,
+      "loss": 2.5031,
+      "step": 5385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012564983118990017,
+      "loss": 2.3688,
+      "step": 5386
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012551051608720776,
+      "loss": 1.8034,
+      "step": 5387
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012537114800525054,
+      "loss": 3.1137,
+      "step": 5388
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012523172723346275,
+      "loss": 5.1297,
+      "step": 5389
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012509225406138807,
+      "loss": 3.4066,
+      "step": 5390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001249527287786792,
+      "loss": 4.3762,
+      "step": 5391
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012481315167509687,
+      "loss": 3.3913,
+      "step": 5392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012467352304050943,
+      "loss": 2.4283,
+      "step": 5393
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001245338431648923,
+      "loss": 4.7096,
+      "step": 5394
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012439411233832734,
+      "loss": 2.0145,
+      "step": 5395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012425433085100227,
+      "loss": 4.2143,
+      "step": 5396
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012411449899320972,
+      "loss": 4.7234,
+      "step": 5397
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012397461705534728,
+      "loss": 3.718,
+      "step": 5398
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001238346853279164,
+      "loss": 2.5066,
+      "step": 5399
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012369470410152194,
+      "loss": 1.7847,
+      "step": 5400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012355467366687156,
+      "loss": 2.2833,
+      "step": 5401
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012341459431477512,
+      "loss": 3.4455,
+      "step": 5402
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001232744663361441,
+      "loss": 5.2422,
+      "step": 5403
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012313429002199094,
+      "loss": 3.3818,
+      "step": 5404
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012299406566342836,
+      "loss": 2.9951,
+      "step": 5405
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012285379355166898,
+      "loss": 4.4397,
+      "step": 5406
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012271347397802454,
+      "loss": 2.9142,
+      "step": 5407
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012257310723390544,
+      "loss": 2.6375,
+      "step": 5408
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012243269361081994,
+      "loss": 3.4468,
+      "step": 5409
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012229223340037368,
+      "loss": 3.1745,
+      "step": 5410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012215172689426904,
+      "loss": 1.7342,
+      "step": 5411
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012201117438430458,
+      "loss": 4.6764,
+      "step": 5412
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012187057616237447,
+      "loss": 2.8161,
+      "step": 5413
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00012172993252046756,
+      "loss": 1.1718,
+      "step": 5414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012158924375066725,
+      "loss": 5.1997,
+      "step": 5415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012144851014515064,
+      "loss": 3.9955,
+      "step": 5416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012130773199618786,
+      "loss": 3.9185,
+      "step": 5417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012116690959614161,
+      "loss": 1.1814,
+      "step": 5418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012102604323746645,
+      "loss": 4.9994,
+      "step": 5419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012088513321270829,
+      "loss": 3.345,
+      "step": 5420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012074417981450365,
+      "loss": 4.6796,
+      "step": 5421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001206031833355791,
+      "loss": 4.2478,
+      "step": 5422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012046214406875087,
+      "loss": 2.0133,
+      "step": 5423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012032106230692388,
+      "loss": 3.1827,
+      "step": 5424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012017993834309131,
+      "loss": 1.7767,
+      "step": 5425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012003877247033404,
+      "loss": 3.5322,
+      "step": 5426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011989756498182002,
+      "loss": 4.9104,
+      "step": 5427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011975631617080339,
+      "loss": 3.96,
+      "step": 5428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011961502633062441,
+      "loss": 3.538,
+      "step": 5429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011947369575470838,
+      "loss": 5.285,
+      "step": 5430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011933232473656524,
+      "loss": 2.2217,
+      "step": 5431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011919091356978893,
+      "loss": 4.055,
+      "step": 5432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011904946254805677,
+      "loss": 5.2614,
+      "step": 5433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011890797196512882,
+      "loss": 6.5961,
+      "step": 5434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001187664421148473,
+      "loss": 3.9701,
+      "step": 5435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011862487329113612,
+      "loss": 2.9062,
+      "step": 5436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001184832657879998,
+      "loss": 2.8162,
+      "step": 5437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011834161989952349,
+      "loss": 4.9796,
+      "step": 5438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011819993591987193,
+      "loss": 2.0845,
+      "step": 5439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011805821414328901,
+      "loss": 3.4697,
+      "step": 5440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011791645486409715,
+      "loss": 3.6732,
+      "step": 5441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011777465837669652,
+      "loss": 2.9764,
+      "step": 5442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001176328249755647,
+      "loss": 3.3593,
+      "step": 5443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011749095495525586,
+      "loss": 2.7487,
+      "step": 5444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011734904861040032,
+      "loss": 2.1969,
+      "step": 5445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011720710623570359,
+      "loss": 3.3076,
+      "step": 5446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011706512812594623,
+      "loss": 4.3397,
+      "step": 5447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000116923114575983,
+      "loss": 4.2856,
+      "step": 5448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011678106588074215,
+      "loss": 4.63,
+      "step": 5449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011663898233522505,
+      "loss": 4.3743,
+      "step": 5450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011649686423450535,
+      "loss": 4.0129,
+      "step": 5451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011635471187372847,
+      "loss": 3.6767,
+      "step": 5452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011621252554811099,
+      "loss": 4.9415,
+      "step": 5453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011607030555294001,
+      "loss": 3.3714,
+      "step": 5454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001159280521835727,
+      "loss": 4.3423,
+      "step": 5455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011578576573543536,
+      "loss": 3.7252,
+      "step": 5456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011564344650402304,
+      "loss": 3.802,
+      "step": 5457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011550109478489887,
+      "loss": 4.6754,
+      "step": 5458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011535871087369357,
+      "loss": 4.8895,
+      "step": 5459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001152162950661044,
+      "loss": 3.9948,
+      "step": 5460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001150738476578952,
+      "loss": 4.3858,
+      "step": 5461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011493136894489527,
+      "loss": 3.7503,
+      "step": 5462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011478885922299896,
+      "loss": 3.6411,
+      "step": 5463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011464631878816501,
+      "loss": 4.6913,
+      "step": 5464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011450374793641594,
+      "loss": 3.6632,
+      "step": 5465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011436114696383749,
+      "loss": 3.4987,
+      "step": 5466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011421851616657793,
+      "loss": 3.7446,
+      "step": 5467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011407585584084732,
+      "loss": 4.54,
+      "step": 5468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011393316628291728,
+      "loss": 3.2563,
+      "step": 5469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011379044778911999,
+      "loss": 4.7622,
+      "step": 5470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011364770065584776,
+      "loss": 3.3135,
+      "step": 5471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011350492517955238,
+      "loss": 4.8041,
+      "step": 5472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011336212165674454,
+      "loss": 4.0978,
+      "step": 5473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011321929038399312,
+      "loss": 4.8156,
+      "step": 5474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011307643165792463,
+      "loss": 1.7122,
+      "step": 5475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011293354577522273,
+      "loss": 3.1119,
+      "step": 5476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011279063303262715,
+      "loss": 3.6845,
+      "step": 5477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011264769372693373,
+      "loss": 3.2134,
+      "step": 5478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011250472815499335,
+      "loss": 2.9211,
+      "step": 5479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011236173661371145,
+      "loss": 3.99,
+      "step": 5480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011221871940004737,
+      "loss": 4.1181,
+      "step": 5481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011207567681101384,
+      "loss": 3.9516,
+      "step": 5482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001119326091436762,
+      "loss": 4.0602,
+      "step": 5483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011178951669515194,
+      "loss": 2.5181,
+      "step": 5484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011164639976261007,
+      "loss": 3.7045,
+      "step": 5485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011150325864326997,
+      "loss": 2.5153,
+      "step": 5486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011136009363440218,
+      "loss": 5.1357,
+      "step": 5487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011121690503332607,
+      "loss": 3.068,
+      "step": 5488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011107369313741027,
+      "loss": 3.358,
+      "step": 5489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011093045824407186,
+      "loss": 4.0057,
+      "step": 5490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011078720065077541,
+      "loss": 4.0282,
+      "step": 5491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011064392065503295,
+      "loss": 3.6656,
+      "step": 5492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011050061855440284,
+      "loss": 3.0703,
+      "step": 5493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011035729464648943,
+      "loss": 4.4825,
+      "step": 5494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011021394922894233,
+      "loss": 3.2514,
+      "step": 5495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00011007058259945582,
+      "loss": 2.4642,
+      "step": 5496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010992719505576826,
+      "loss": 2.1637,
+      "step": 5497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001097837868956614,
+      "loss": 2.7467,
+      "step": 5498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010964035841695992,
+      "loss": 3.3566,
+      "step": 5499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010949690991753041,
+      "loss": 5.539,
+      "step": 5500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010935344169528136,
+      "loss": 3.636,
+      "step": 5501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010920995404816205,
+      "loss": 2.9384,
+      "step": 5502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010906644727416218,
+      "loss": 1.7699,
+      "step": 5503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010892292167131109,
+      "loss": 5.2123,
+      "step": 5504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001087793775376773,
+      "loss": 1.4921,
+      "step": 5505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010863581517136776,
+      "loss": 1.7406,
+      "step": 5506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010849223487052733,
+      "loss": 6.1473,
+      "step": 5507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010834863693333814,
+      "loss": 3.4387,
+      "step": 5508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010820502165801877,
+      "loss": 4.3485,
+      "step": 5509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000108061389342824,
+      "loss": 4.4457,
+      "step": 5510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010791774028604391,
+      "loss": 2.6966,
+      "step": 5511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010777407478600341,
+      "loss": 5.1003,
+      "step": 5512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010763039314106152,
+      "loss": 3.8427,
+      "step": 5513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010748669564961073,
+      "loss": 4.4372,
+      "step": 5514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010734298261007659,
+      "loss": 4.4111,
+      "step": 5515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010719925432091676,
+      "loss": 2.8079,
+      "step": 5516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001070555110806208,
+      "loss": 4.54,
+      "step": 5517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010691175318770881,
+      "loss": 1.4158,
+      "step": 5518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010676798094073213,
+      "loss": 3.1458,
+      "step": 5519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001066241946382713,
+      "loss": 2.1782,
+      "step": 5520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001064803945789362,
+      "loss": 4.1587,
+      "step": 5521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001063365810613654,
+      "loss": 4.9544,
+      "step": 5522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010619275438422511,
+      "loss": 3.7955,
+      "step": 5523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001060489148462092,
+      "loss": 4.3905,
+      "step": 5524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010590506274603813,
+      "loss": 4.9464,
+      "step": 5525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010576119838245842,
+      "loss": 3.8898,
+      "step": 5526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010561732205424208,
+      "loss": 3.2374,
+      "step": 5527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000105473434060186,
+      "loss": 4.7712,
+      "step": 5528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010532953469911125,
+      "loss": 2.6548,
+      "step": 5529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010518562426986249,
+      "loss": 3.4113,
+      "step": 5530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010504170307130754,
+      "loss": 3.1742,
+      "step": 5531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010489777140233625,
+      "loss": 0.9218,
+      "step": 5532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010475382956186048,
+      "loss": 3.8299,
+      "step": 5533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010460987784881317,
+      "loss": 2.5356,
+      "step": 5534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010446591656214772,
+      "loss": 4.8955,
+      "step": 5535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010432194600083742,
+      "loss": 5.6838,
+      "step": 5536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010417796646387486,
+      "loss": 3.9082,
+      "step": 5537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010403397825027122,
+      "loss": 2.2672,
+      "step": 5538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001038899816590558,
+      "loss": 4.6001,
+      "step": 5539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010374597698927503,
+      "loss": 4.0564,
+      "step": 5540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010360196453999245,
+      "loss": 1.4487,
+      "step": 5541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010345794461028755,
+      "loss": 5.1357,
+      "step": 5542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010331391749925545,
+      "loss": 3.1433,
+      "step": 5543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010316988350600614,
+      "loss": 4.6478,
+      "step": 5544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001030258429296639,
+      "loss": 3.6029,
+      "step": 5545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010288179606936672,
+      "loss": 2.105,
+      "step": 5546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001027377432242656,
+      "loss": 1.5399,
+      "step": 5547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010259368469352404,
+      "loss": 5.0061,
+      "step": 5548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010244962077631717,
+      "loss": 4.234,
+      "step": 5549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001023055517718313,
+      "loss": 1.5645,
+      "step": 5550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010216147797926383,
+      "loss": 5.0323,
+      "step": 5551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010201739969782147,
+      "loss": 3.3444,
+      "step": 5552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010187331722672062,
+      "loss": 3.4442,
+      "step": 5553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010172923086518606,
+      "loss": 2.6519,
+      "step": 5554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010158514091245092,
+      "loss": 3.2137,
+      "step": 5555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010144104766775572,
+      "loss": 4.5934,
+      "step": 5556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010129695143034781,
+      "loss": 2.7312,
+      "step": 5557
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001011528524994807,
+      "loss": 3.0099,
+      "step": 5558
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010100875117441363,
+      "loss": 5.2993,
+      "step": 5559
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010086464775441062,
+      "loss": 3.8233,
+      "step": 5560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010072054253874024,
+      "loss": 4.9942,
+      "step": 5561
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010057643582667471,
+      "loss": 4.6087,
+      "step": 5562
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010043232791748923,
+      "loss": 2.8269,
+      "step": 5563
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010028821911046163,
+      "loss": 3.7981,
+      "step": 5564
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010014410970487158,
+      "loss": 3.2126,
+      "step": 5565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010000000000000003,
+      "loss": 5.4441,
+      "step": 5566
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.985589029512848e-05,
+      "loss": 3.9594,
+      "step": 5567
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.971178088953845e-05,
+      "loss": 3.3012,
+      "step": 5568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.956767208251085e-05,
+      "loss": 4.8113,
+      "step": 5569
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.942356417332535e-05,
+      "loss": 1.3955,
+      "step": 5570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.927945746125984e-05,
+      "loss": 2.6099,
+      "step": 5571
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.913535224558945e-05,
+      "loss": 3.6693,
+      "step": 5572
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.899124882558645e-05,
+      "loss": 4.5581,
+      "step": 5573
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.884714750051936e-05,
+      "loss": 6.1733,
+      "step": 5574
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.870304856965227e-05,
+      "loss": 4.8061,
+      "step": 5575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.855895233224434e-05,
+      "loss": 1.4784,
+      "step": 5576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.841485908754916e-05,
+      "loss": 3.1093,
+      "step": 5577
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.827076913481402e-05,
+      "loss": 2.7161,
+      "step": 5578
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.812668277327946e-05,
+      "loss": 3.8907,
+      "step": 5579
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.79826003021786e-05,
+      "loss": 3.3913,
+      "step": 5580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.783852202073624e-05,
+      "loss": 5.8769,
+      "step": 5581
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.769444822816875e-05,
+      "loss": 3.9217,
+      "step": 5582
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.755037922368289e-05,
+      "loss": 3.1719,
+      "step": 5583
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.740631530647602e-05,
+      "loss": 3.4813,
+      "step": 5584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.726225677573448e-05,
+      "loss": 3.3804,
+      "step": 5585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.711820393063335e-05,
+      "loss": 3.4843,
+      "step": 5586
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.697415707033617e-05,
+      "loss": 3.1884,
+      "step": 5587
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.683011649399394e-05,
+      "loss": 2.8408,
+      "step": 5588
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.668608250074463e-05,
+      "loss": 1.9692,
+      "step": 5589
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.654205538971252e-05,
+      "loss": 3.8228,
+      "step": 5590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.639803546000762e-05,
+      "loss": 3.1275,
+      "step": 5591
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.625402301072503e-05,
+      "loss": 2.6927,
+      "step": 5592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.611001834094427e-05,
+      "loss": 3.8915,
+      "step": 5593
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.596602174972886e-05,
+      "loss": 3.5443,
+      "step": 5594
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.582203353612522e-05,
+      "loss": 4.2841,
+      "step": 5595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.567805399916265e-05,
+      "loss": 3.5272,
+      "step": 5596
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.553408343785235e-05,
+      "loss": 3.2564,
+      "step": 5597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.53901221511869e-05,
+      "loss": 3.6711,
+      "step": 5598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.52461704381396e-05,
+      "loss": 2.2848,
+      "step": 5599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.510222859766383e-05,
+      "loss": 3.9135,
+      "step": 5600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.495829692869254e-05,
+      "loss": 3.8639,
+      "step": 5601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.481437573013759e-05,
+      "loss": 4.7719,
+      "step": 5602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.467046530088883e-05,
+      "loss": 3.2916,
+      "step": 5603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.452656593981406e-05,
+      "loss": 2.9318,
+      "step": 5604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.438267794575798e-05,
+      "loss": 4.132,
+      "step": 5605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.423880161754165e-05,
+      "loss": 3.0521,
+      "step": 5606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.409493725396195e-05,
+      "loss": 3.1693,
+      "step": 5607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.395108515379086e-05,
+      "loss": 3.3401,
+      "step": 5608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.380724561577497e-05,
+      "loss": 2.8139,
+      "step": 5609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.366341893863468e-05,
+      "loss": 3.2937,
+      "step": 5610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.351960542106388e-05,
+      "loss": 3.8979,
+      "step": 5611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.337580536172877e-05,
+      "loss": 2.2771,
+      "step": 5612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.323201905926794e-05,
+      "loss": 4.1266,
+      "step": 5613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.308824681229125e-05,
+      "loss": 3.2596,
+      "step": 5614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.294448891937926e-05,
+      "loss": 3.6044,
+      "step": 5615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.28007456790833e-05,
+      "loss": 1.4646,
+      "step": 5616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.265701738992349e-05,
+      "loss": 4.6085,
+      "step": 5617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.251330435038933e-05,
+      "loss": 2.2269,
+      "step": 5618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.236960685893856e-05,
+      "loss": 2.5445,
+      "step": 5619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.222592521399665e-05,
+      "loss": 4.5066,
+      "step": 5620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.208225971395616e-05,
+      "loss": 3.4294,
+      "step": 5621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.193861065717608e-05,
+      "loss": 4.3222,
+      "step": 5622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.17949783419813e-05,
+      "loss": 3.282,
+      "step": 5623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.165136306666192e-05,
+      "loss": 3.0066,
+      "step": 5624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.150776512947275e-05,
+      "loss": 3.4138,
+      "step": 5625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.136418482863231e-05,
+      "loss": 1.877,
+      "step": 5626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.122062246232277e-05,
+      "loss": 2.9548,
+      "step": 5627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.107707832868898e-05,
+      "loss": 5.6738,
+      "step": 5628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.09335527258379e-05,
+      "loss": 3.9868,
+      "step": 5629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.079004595183801e-05,
+      "loss": 3.7319,
+      "step": 5630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.064655830471871e-05,
+      "loss": 3.481,
+      "step": 5631
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.050309008246965e-05,
+      "loss": 2.9829,
+      "step": 5632
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.035964158304016e-05,
+      "loss": 2.5347,
+      "step": 5633
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.021621310433868e-05,
+      "loss": 5.1191,
+      "step": 5634
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.007280494423181e-05,
+      "loss": 4.0436,
+      "step": 5635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.992941740054425e-05,
+      "loss": 4.6252,
+      "step": 5636
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.978605077105775e-05,
+      "loss": 3.6014,
+      "step": 5637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.964270535351065e-05,
+      "loss": 4.0038,
+      "step": 5638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.949938144559724e-05,
+      "loss": 3.4097,
+      "step": 5639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.935607934496712e-05,
+      "loss": 0.831,
+      "step": 5640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.921279934922466e-05,
+      "loss": 4.0215,
+      "step": 5641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.906954175592822e-05,
+      "loss": 4.2341,
+      "step": 5642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.89263068625898e-05,
+      "loss": 1.9985,
+      "step": 5643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.8783094966674e-05,
+      "loss": 1.1326,
+      "step": 5644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.86399063655979e-05,
+      "loss": 3.1922,
+      "step": 5645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.849674135673009e-05,
+      "loss": 4.4856,
+      "step": 5646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.835360023739001e-05,
+      "loss": 4.4926,
+      "step": 5647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.821048330484814e-05,
+      "loss": 3.7006,
+      "step": 5648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.806739085632388e-05,
+      "loss": 3.1484,
+      "step": 5649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.792432318898623e-05,
+      "loss": 1.7988,
+      "step": 5650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.778128059995269e-05,
+      "loss": 3.8663,
+      "step": 5651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.763826338628863e-05,
+      "loss": 3.8821,
+      "step": 5652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.749527184500671e-05,
+      "loss": 2.7814,
+      "step": 5653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.735230627306633e-05,
+      "loss": 3.7347,
+      "step": 5654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.720936696737291e-05,
+      "loss": 2.1295,
+      "step": 5655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.706645422477735e-05,
+      "loss": 1.0188,
+      "step": 5656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.692356834207544e-05,
+      "loss": 3.1181,
+      "step": 5657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.678070961600695e-05,
+      "loss": 2.0098,
+      "step": 5658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.663787834325553e-05,
+      "loss": 3.2446,
+      "step": 5659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.649507482044768e-05,
+      "loss": 1.3846,
+      "step": 5660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.635229934415232e-05,
+      "loss": 3.1674,
+      "step": 5661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.620955221088009e-05,
+      "loss": 2.8632,
+      "step": 5662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.60668337170828e-05,
+      "loss": 2.4158,
+      "step": 5663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.592414415915273e-05,
+      "loss": 3.3156,
+      "step": 5664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.578148383342214e-05,
+      "loss": 4.1942,
+      "step": 5665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.563885303616259e-05,
+      "loss": 0.8172,
+      "step": 5666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.549625206358413e-05,
+      "loss": 4.938,
+      "step": 5667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.535368121183498e-05,
+      "loss": 4.3131,
+      "step": 5668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.521114077700103e-05,
+      "loss": 3.3248,
+      "step": 5669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.506863105510489e-05,
+      "loss": 1.5731,
+      "step": 5670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.492615234210496e-05,
+      "loss": 3.8352,
+      "step": 5671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.478370493389575e-05,
+      "loss": 3.5706,
+      "step": 5672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.464128912630659e-05,
+      "loss": 3.4681,
+      "step": 5673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.44989052151012e-05,
+      "loss": 2.5602,
+      "step": 5674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.435655349597704e-05,
+      "loss": 2.133,
+      "step": 5675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.421423426456471e-05,
+      "loss": 3.2811,
+      "step": 5676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.407194781642738e-05,
+      "loss": 2.4266,
+      "step": 5677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.392969444706007e-05,
+      "loss": 4.7839,
+      "step": 5678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.378747445188898e-05,
+      "loss": 3.0872,
+      "step": 5679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.364528812627153e-05,
+      "loss": 3.4402,
+      "step": 5680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.350313576549465e-05,
+      "loss": 2.9748,
+      "step": 5681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.336101766477493e-05,
+      "loss": 4.2048,
+      "step": 5682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.321893411925782e-05,
+      "loss": 2.4749,
+      "step": 5683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.307688542401717e-05,
+      "loss": 2.52,
+      "step": 5684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.293487187405393e-05,
+      "loss": 4.3217,
+      "step": 5685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.279289376429657e-05,
+      "loss": 3.5208,
+      "step": 5686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.265095138959984e-05,
+      "loss": 5.3406,
+      "step": 5687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.250904504474421e-05,
+      "loss": 3.9533,
+      "step": 5688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.236717502443537e-05,
+      "loss": 4.4469,
+      "step": 5689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.222534162330356e-05,
+      "loss": 3.0943,
+      "step": 5690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.208354513590292e-05,
+      "loss": 3.7718,
+      "step": 5691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.194178585671104e-05,
+      "loss": 1.6704,
+      "step": 5692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.180006408012815e-05,
+      "loss": 4.6448,
+      "step": 5693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.165838010047659e-05,
+      "loss": 4.2533,
+      "step": 5694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.151673421200028e-05,
+      "loss": 3.7315,
+      "step": 5695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.137512670886394e-05,
+      "loss": 4.6147,
+      "step": 5696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.123355788515268e-05,
+      "loss": 4.0328,
+      "step": 5697
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.109202803487117e-05,
+      "loss": 4.6394,
+      "step": 5698
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.095053745194321e-05,
+      "loss": 3.9047,
+      "step": 5699
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.080908643021104e-05,
+      "loss": 4.8046,
+      "step": 5700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.066767526343492e-05,
+      "loss": 3.4142,
+      "step": 5701
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.052630424529178e-05,
+      "loss": 3.2665,
+      "step": 5702
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.038497366937575e-05,
+      "loss": 0.5254,
+      "step": 5703
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.024368382919676e-05,
+      "loss": 2.4317,
+      "step": 5704
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.010243501818014e-05,
+      "loss": 3.8869,
+      "step": 5705
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.996122752966604e-05,
+      "loss": 2.6196,
+      "step": 5706
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.982006165690876e-05,
+      "loss": 3.4457,
+      "step": 5707
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.96789376930762e-05,
+      "loss": 4.1179,
+      "step": 5708
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.95378559312492e-05,
+      "loss": 3.1208,
+      "step": 5709
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.939681666442098e-05,
+      "loss": 3.5072,
+      "step": 5710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.925582018549633e-05,
+      "loss": 4.2613,
+      "step": 5711
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.91148667872917e-05,
+      "loss": 4.3009,
+      "step": 5712
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.897395676253352e-05,
+      "loss": 2.3749,
+      "step": 5713
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.883309040385837e-05,
+      "loss": 4.638,
+      "step": 5714
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.869226800381231e-05,
+      "loss": 3.021,
+      "step": 5715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.855148985484953e-05,
+      "loss": 3.951,
+      "step": 5716
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.84107562493329e-05,
+      "loss": 3.5513,
+      "step": 5717
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.827006747953259e-05,
+      "loss": 5.0401,
+      "step": 5718
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.812942383762566e-05,
+      "loss": 3.6699,
+      "step": 5719
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.798882561569548e-05,
+      "loss": 1.3296,
+      "step": 5720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.784827310573103e-05,
+      "loss": 2.2581,
+      "step": 5721
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.77077665996264e-05,
+      "loss": 2.9865,
+      "step": 5722
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.756730638918012e-05,
+      "loss": 3.4473,
+      "step": 5723
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.742689276609461e-05,
+      "loss": 3.3926,
+      "step": 5724
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.728652602197551e-05,
+      "loss": 5.6973,
+      "step": 5725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.714620644833111e-05,
+      "loss": 3.4892,
+      "step": 5726
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.700593433657172e-05,
+      "loss": 2.4708,
+      "step": 5727
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.686570997800911e-05,
+      "loss": 2.7115,
+      "step": 5728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.672553366385589e-05,
+      "loss": 0.9883,
+      "step": 5729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.658540568522483e-05,
+      "loss": 3.6127,
+      "step": 5730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.644532633312841e-05,
+      "loss": 2.5587,
+      "step": 5731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.630529589847804e-05,
+      "loss": 2.425,
+      "step": 5732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.616531467208376e-05,
+      "loss": 2.7636,
+      "step": 5733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.602538294465287e-05,
+      "loss": 4.0515,
+      "step": 5734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.588550100679044e-05,
+      "loss": 4.233,
+      "step": 5735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.57456691489979e-05,
+      "loss": 4.5607,
+      "step": 5736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.560588766167271e-05,
+      "loss": 3.3437,
+      "step": 5737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.546615683510777e-05,
+      "loss": 3.6766,
+      "step": 5738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.532647695949064e-05,
+      "loss": 1.9198,
+      "step": 5739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.51868483249032e-05,
+      "loss": 2.9665,
+      "step": 5740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.504727122132088e-05,
+      "loss": 3.62,
+      "step": 5741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.4907745938612e-05,
+      "loss": 2.5871,
+      "step": 5742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.476827276653723e-05,
+      "loss": 3.9261,
+      "step": 5743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.462885199474945e-05,
+      "loss": 3.1297,
+      "step": 5744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.448948391279221e-05,
+      "loss": 4.1396,
+      "step": 5745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.435016881009981e-05,
+      "loss": 3.4951,
+      "step": 5746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.421090697599673e-05,
+      "loss": 2.0667,
+      "step": 5747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.407169869969626e-05,
+      "loss": 1.7521,
+      "step": 5748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.393254427030101e-05,
+      "loss": 2.6155,
+      "step": 5749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.379344397680157e-05,
+      "loss": 1.4222,
+      "step": 5750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.365439810807611e-05,
+      "loss": 5.5235,
+      "step": 5751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.351540695288979e-05,
+      "loss": 3.591,
+      "step": 5752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.337647079989412e-05,
+      "loss": 2.5905,
+      "step": 5753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.32375899376264e-05,
+      "loss": 2.5899,
+      "step": 5754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.309876465450909e-05,
+      "loss": 2.5796,
+      "step": 5755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.295999523884924e-05,
+      "loss": 1.7404,
+      "step": 5756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.282128197883785e-05,
+      "loss": 3.618,
+      "step": 5757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.268262516254934e-05,
+      "loss": 3.831,
+      "step": 5758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.254402507794089e-05,
+      "loss": 1.208,
+      "step": 5759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.240548201285184e-05,
+      "loss": 2.8676,
+      "step": 5760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.226699625500312e-05,
+      "loss": 2.2912,
+      "step": 5761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.21285680919967e-05,
+      "loss": 3.0577,
+      "step": 5762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.199019781131489e-05,
+      "loss": 3.1668,
+      "step": 5763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.185188570031978e-05,
+      "loss": 4.0763,
+      "step": 5764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.171363204625283e-05,
+      "loss": 3.9535,
+      "step": 5765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.157543713623366e-05,
+      "loss": 3.0072,
+      "step": 5766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.14373012572603e-05,
+      "loss": 3.0836,
+      "step": 5767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.129922469620809e-05,
+      "loss": 4.1132,
+      "step": 5768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.116120773982914e-05,
+      "loss": 3.8476,
+      "step": 5769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.102325067475179e-05,
+      "loss": 2.0228,
+      "step": 5770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.088535378748002e-05,
+      "loss": 0.488,
+      "step": 5771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.07475173643928e-05,
+      "loss": 1.9256,
+      "step": 5772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.060974169174355e-05,
+      "loss": 1.8927,
+      "step": 5773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.04720270556595e-05,
+      "loss": 1.3781,
+      "step": 5774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.0334373742141e-05,
+      "loss": 3.2742,
+      "step": 5775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.019678203706154e-05,
+      "loss": 2.8495,
+      "step": 5776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.005925222616612e-05,
+      "loss": 2.5362,
+      "step": 5777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 6.992178459507157e-05,
+      "loss": 3.5808,
+      "step": 5778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 6.978437942926507e-05,
+      "loss": 4.1004,
+      "step": 5779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 6.96470370141046e-05,
+      "loss": 4.0652,
+      "step": 5780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 6.950975763481762e-05,
+      "loss": 2.6863,
+      "step": 5781
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.937254157650073e-05,
+      "loss": 3.1985,
+      "step": 5782
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.923538912411895e-05,
+      "loss": 2.1266,
+      "step": 5783
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.909830056250529e-05,
+      "loss": 2.3151,
+      "step": 5784
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.896127617636003e-05,
+      "loss": 3.5983,
+      "step": 5785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.882431625025015e-05,
+      "loss": 2.1752,
+      "step": 5786
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.868742106860883e-05,
+      "loss": 4.0267,
+      "step": 5787
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.855059091573472e-05,
+      "loss": 3.7479,
+      "step": 5788
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.841382607579148e-05,
+      "loss": 3.5264,
+      "step": 5789
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.827712683280707e-05,
+      "loss": 4.5087,
+      "step": 5790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.814049347067328e-05,
+      "loss": 3.826,
+      "step": 5791
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.800392627314504e-05,
+      "loss": 1.6947,
+      "step": 5792
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.786742552383988e-05,
+      "loss": 3.4003,
+      "step": 5793
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.773099150623739e-05,
+      "loss": 4.2304,
+      "step": 5794
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.759462450367844e-05,
+      "loss": 4.9625,
+      "step": 5795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.745832479936505e-05,
+      "loss": 2.6775,
+      "step": 5796
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.732209267635886e-05,
+      "loss": 6.3136,
+      "step": 5797
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.718592841758172e-05,
+      "loss": 3.3696,
+      "step": 5798
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.704983230581436e-05,
+      "loss": 4.2365,
+      "step": 5799
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.691380462369597e-05,
+      "loss": 3.4909,
+      "step": 5800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.677784565372366e-05,
+      "loss": 2.7353,
+      "step": 5801
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.664195567825181e-05,
+      "loss": 2.0349,
+      "step": 5802
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.650613497949154e-05,
+      "loss": 3.5496,
+      "step": 5803
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.637038383951007e-05,
+      "loss": 4.048,
+      "step": 5804
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.623470254023019e-05,
+      "loss": 4.7727,
+      "step": 5805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.609909136342962e-05,
+      "loss": 4.339,
+      "step": 5806
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.596355059074034e-05,
+      "loss": 1.5886,
+      "step": 5807
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.582808050364854e-05,
+      "loss": 1.5785,
+      "step": 5808
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.569268138349317e-05,
+      "loss": 4.0622,
+      "step": 5809
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.555735351146612e-05,
+      "loss": 3.1834,
+      "step": 5810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.54220971686108e-05,
+      "loss": 2.8512,
+      "step": 5811
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.528691263582258e-05,
+      "loss": 2.7128,
+      "step": 5812
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.515180019384754e-05,
+      "loss": 4.7674,
+      "step": 5813
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.501676012328204e-05,
+      "loss": 1.3461,
+      "step": 5814
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.488179270457211e-05,
+      "loss": 3.3172,
+      "step": 5815
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.474689821801297e-05,
+      "loss": 2.6448,
+      "step": 5816
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.461207694374827e-05,
+      "loss": 2.6898,
+      "step": 5817
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.447732916176977e-05,
+      "loss": 3.8276,
+      "step": 5818
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.434265515191642e-05,
+      "loss": 3.8879,
+      "step": 5819
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.420805519387412e-05,
+      "loss": 3.8882,
+      "step": 5820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.40735295671749e-05,
+      "loss": 3.1491,
+      "step": 5821
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.393907855119643e-05,
+      "loss": 3.1642,
+      "step": 5822
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.380470242516145e-05,
+      "loss": 5.5097,
+      "step": 5823
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.367040146813717e-05,
+      "loss": 2.5979,
+      "step": 5824
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.353617595903467e-05,
+      "loss": 2.5081,
+      "step": 5825
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.340202617660838e-05,
+      "loss": 2.3105,
+      "step": 5826
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.326795239945542e-05,
+      "loss": 2.159,
+      "step": 5827
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.313395490601526e-05,
+      "loss": 4.7427,
+      "step": 5828
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.300003397456842e-05,
+      "loss": 5.7737,
+      "step": 5829
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.286618988323699e-05,
+      "loss": 3.7156,
+      "step": 5830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.273242290998322e-05,
+      "loss": 4.4617,
+      "step": 5831
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.259873333260926e-05,
+      "loss": 2.1095,
+      "step": 5832
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.246512142875651e-05,
+      "loss": 3.7442,
+      "step": 5833
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.233158747590508e-05,
+      "loss": 2.7415,
+      "step": 5834
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.219813175137316e-05,
+      "loss": 4.4172,
+      "step": 5835
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.206475453231653e-05,
+      "loss": 2.4911,
+      "step": 5836
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.193145609572786e-05,
+      "loss": 4.0837,
+      "step": 5837
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.179823671843627e-05,
+      "loss": 4.2414,
+      "step": 5838
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.166509667710652e-05,
+      "loss": 4.7073,
+      "step": 5839
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.153203624823905e-05,
+      "loss": 1.9264,
+      "step": 5840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.139905570816855e-05,
+      "loss": 3.2637,
+      "step": 5841
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.126615533306407e-05,
+      "loss": 2.4439,
+      "step": 5842
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.113333539892769e-05,
+      "loss": 2.634,
+      "step": 5843
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.100059618159497e-05,
+      "loss": 4.9527,
+      "step": 5844
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.086793795673361e-05,
+      "loss": 5.0117,
+      "step": 5845
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.073536099984316e-05,
+      "loss": 4.9842,
+      "step": 5846
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.0602865586254366e-05,
+      "loss": 4.4576,
+      "step": 5847
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.047045199112864e-05,
+      "loss": 4.4695,
+      "step": 5848
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.0338120489457486e-05,
+      "loss": 3.9614,
+      "step": 5849
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.020587135606191e-05,
+      "loss": 3.5151,
+      "step": 5850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 6.007370486559185e-05,
+      "loss": 2.2908,
+      "step": 5851
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.994162129252562e-05,
+      "loss": 4.1372,
+      "step": 5852
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.980962091116934e-05,
+      "loss": 1.2491,
+      "step": 5853
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.967770399565635e-05,
+      "loss": 3.0221,
+      "step": 5854
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.954587081994668e-05,
+      "loss": 3.3849,
+      "step": 5855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.9414121657826415e-05,
+      "loss": 3.1262,
+      "step": 5856
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.928245678290718e-05,
+      "loss": 2.6778,
+      "step": 5857
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.915087646862556e-05,
+      "loss": 2.3316,
+      "step": 5858
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.9019380988242675e-05,
+      "loss": 2.9169,
+      "step": 5859
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.8887970614843e-05,
+      "loss": 2.3224,
+      "step": 5860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.875664562133474e-05,
+      "loss": 4.1771,
+      "step": 5861
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.862540628044862e-05,
+      "loss": 3.4531,
+      "step": 5862
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.849425286473746e-05,
+      "loss": 2.781,
+      "step": 5863
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.836318564657571e-05,
+      "loss": 3.2235,
+      "step": 5864
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.823220489815872e-05,
+      "loss": 3.7612,
+      "step": 5865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.810131089150234e-05,
+      "loss": 3.2481,
+      "step": 5866
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.797050389844221e-05,
+      "loss": 1.8505,
+      "step": 5867
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.7839784190633294e-05,
+      "loss": 1.11,
+      "step": 5868
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.77091520395493e-05,
+      "loss": 4.1303,
+      "step": 5869
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.757860771648207e-05,
+      "loss": 3.6341,
+      "step": 5870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.744815149254093e-05,
+      "loss": 1.9059,
+      "step": 5871
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.731778363865268e-05,
+      "loss": 2.7231,
+      "step": 5872
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.71875044255603e-05,
+      "loss": 3.1668,
+      "step": 5873
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.705731412382237e-05,
+      "loss": 3.627,
+      "step": 5874
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.692721300381328e-05,
+      "loss": 4.3642,
+      "step": 5875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.6797201335722086e-05,
+      "loss": 3.692,
+      "step": 5876
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.6667279389552005e-05,
+      "loss": 3.6547,
+      "step": 5877
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.653744743512002e-05,
+      "loss": 1.4188,
+      "step": 5878
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.640770574205616e-05,
+      "loss": 3.7082,
+      "step": 5879
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.6278054579803e-05,
+      "loss": 4.7223,
+      "step": 5880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.6148494217615114e-05,
+      "loss": 3.5315,
+      "step": 5881
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.6019024924558525e-05,
+      "loss": 2.4488,
+      "step": 5882
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.588964696951011e-05,
+      "loss": 4.4712,
+      "step": 5883
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.576036062115708e-05,
+      "loss": 3.7228,
+      "step": 5884
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.563116614799635e-05,
+      "loss": 2.856,
+      "step": 5885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.550206381833407e-05,
+      "loss": 2.8282,
+      "step": 5886
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.5373053900285046e-05,
+      "loss": 1.1655,
+      "step": 5887
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.524413666177213e-05,
+      "loss": 4.0632,
+      "step": 5888
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.5115312370525717e-05,
+      "loss": 2.5051,
+      "step": 5889
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.498658129408316e-05,
+      "loss": 4.8146,
+      "step": 5890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.4857943699788426e-05,
+      "loss": 3.8134,
+      "step": 5891
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.472939985479082e-05,
+      "loss": 2.8053,
+      "step": 5892
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.460095002604544e-05,
+      "loss": 2.995,
+      "step": 5893
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.447259448031198e-05,
+      "loss": 2.5807,
+      "step": 5894
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.4344333484154354e-05,
+      "loss": 4.11,
+      "step": 5895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.421616730394006e-05,
+      "loss": 2.4535,
+      "step": 5896
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.4088096205839764e-05,
+      "loss": 4.4132,
+      "step": 5897
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.396012045582661e-05,
+      "loss": 3.3586,
+      "step": 5898
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.383224031967573e-05,
+      "loss": 2.4001,
+      "step": 5899
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.370445606296373e-05,
+      "loss": 2.9183,
+      "step": 5900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.357676795106805e-05,
+      "loss": 1.2053,
+      "step": 5901
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.34491762491665e-05,
+      "loss": 3.7671,
+      "step": 5902
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.3321681222236464e-05,
+      "loss": 4.8538,
+      "step": 5903
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.319428313505508e-05,
+      "loss": 3.7985,
+      "step": 5904
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.306698225219783e-05,
+      "loss": 2.2378,
+      "step": 5905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.293977883803799e-05,
+      "loss": 2.9855,
+      "step": 5906
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.281267315674692e-05,
+      "loss": 4.1475,
+      "step": 5907
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.2685665472292826e-05,
+      "loss": 0.4565,
+      "step": 5908
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.255875604844043e-05,
+      "loss": 4.0318,
+      "step": 5909
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.243194514875035e-05,
+      "loss": 1.9809,
+      "step": 5910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.2305233036578636e-05,
+      "loss": 2.1425,
+      "step": 5911
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.217861997507617e-05,
+      "loss": 2.3218,
+      "step": 5912
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.20521062271881e-05,
+      "loss": 3.0825,
+      "step": 5913
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.192569205565334e-05,
+      "loss": 3.7273,
+      "step": 5914
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.179937772300403e-05,
+      "loss": 3.1686,
+      "step": 5915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.167316349156492e-05,
+      "loss": 3.8482,
+      "step": 5916
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.154704962345292e-05,
+      "loss": 3.9422,
+      "step": 5917
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.1421036380576504e-05,
+      "loss": 1.9999,
+      "step": 5918
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.1295124024635135e-05,
+      "loss": 2.3429,
+      "step": 5919
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.1169312817118784e-05,
+      "loss": 3.5649,
+      "step": 5920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.104360301930737e-05,
+      "loss": 3.2028,
+      "step": 5921
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.091799489227036e-05,
+      "loss": 1.8904,
+      "step": 5922
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.079248869686558e-05,
+      "loss": 1.2601,
+      "step": 5923
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.0667084693739666e-05,
+      "loss": 1.5609,
+      "step": 5924
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.054178314332685e-05,
+      "loss": 1.5908,
+      "step": 5925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.0416584305848626e-05,
+      "loss": 3.7135,
+      "step": 5926
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.029148844131313e-05,
+      "loss": 3.6439,
+      "step": 5927
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.016649580951468e-05,
+      "loss": 1.3205,
+      "step": 5928
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.004160667003319e-05,
+      "loss": 2.4401,
+      "step": 5929
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.9916821282233676e-05,
+      "loss": 2.1066,
+      "step": 5930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.979213990526564e-05,
+      "loss": 4.9306,
+      "step": 5931
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.9667562798062606e-05,
+      "loss": 2.3217,
+      "step": 5932
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.9543090219341556e-05,
+      "loss": 3.042,
+      "step": 5933
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.941872242760237e-05,
+      "loss": 1.4619,
+      "step": 5934
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.9294459681127194e-05,
+      "loss": 0.6845,
+      "step": 5935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.917030223798058e-05,
+      "loss": 3.8636,
+      "step": 5936
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.904625035600753e-05,
+      "loss": 1.9065,
+      "step": 5937
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.8922304292834396e-05,
+      "loss": 4.2275,
+      "step": 5938
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.879846430586761e-05,
+      "loss": 3.5944,
+      "step": 5939
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.867473065229334e-05,
+      "loss": 2.5398,
+      "step": 5940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.855110358907691e-05,
+      "loss": 3.9822,
+      "step": 5941
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.842758337296227e-05,
+      "loss": 2.8722,
+      "step": 5942
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.830417026047149e-05,
+      "loss": 3.1962,
+      "step": 5943
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.818086450790421e-05,
+      "loss": 1.1147,
+      "step": 5944
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.8057666371337095e-05,
+      "loss": 2.346,
+      "step": 5945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.793457610662333e-05,
+      "loss": 3.0145,
+      "step": 5946
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.781159396939206e-05,
+      "loss": 3.8824,
+      "step": 5947
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.768872021504792e-05,
+      "loss": 5.5386,
+      "step": 5948
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.75659550987704e-05,
+      "loss": 3.1851,
+      "step": 5949
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.744329887551341e-05,
+      "loss": 2.6105,
+      "step": 5950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.732075180000472e-05,
+      "loss": 4.3683,
+      "step": 5951
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.719831412674538e-05,
+      "loss": 1.7255,
+      "step": 5952
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.707598611000929e-05,
+      "loss": 4.2561,
+      "step": 5953
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.6953768003842755e-05,
+      "loss": 3.2745,
+      "step": 5954
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.6831660062063356e-05,
+      "loss": 4.4645,
+      "step": 5955
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.670966253826037e-05,
+      "loss": 2.3343,
+      "step": 5956
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.658777568579356e-05,
+      "loss": 2.3975,
+      "step": 5957
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.646599975779291e-05,
+      "loss": 5.3546,
+      "step": 5958
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.6344335007158e-05,
+      "loss": 2.362,
+      "step": 5959
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.6222781686557536e-05,
+      "loss": 5.6064,
+      "step": 5960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.610134004842881e-05,
+      "loss": 1.9834,
+      "step": 5961
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.598001034497718e-05,
+      "loss": 2.4541,
+      "step": 5962
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.585879282817552e-05,
+      "loss": 3.7176,
+      "step": 5963
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.573768774976377e-05,
+      "loss": 2.9174,
+      "step": 5964
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.5616695361248296e-05,
+      "loss": 1.955,
+      "step": 5965
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.5495815913901475e-05,
+      "loss": 2.0906,
+      "step": 5966
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.537504965876098e-05,
+      "loss": 2.1488,
+      "step": 5967
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.525439684663e-05,
+      "loss": 3.7016,
+      "step": 5968
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.513385772807522e-05,
+      "loss": 3.7878,
+      "step": 5969
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.501343255342786e-05,
+      "loss": 3.1859,
+      "step": 5970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.489312157278229e-05,
+      "loss": 4.0651,
+      "step": 5971
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.4772925035995764e-05,
+      "loss": 3.0641,
+      "step": 5972
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.4652843192687864e-05,
+      "loss": 2.718,
+      "step": 5973
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.4532876292239956e-05,
+      "loss": 2.7154,
+      "step": 5974
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.441302458379474e-05,
+      "loss": 3.8096,
+      "step": 5975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.429328831625563e-05,
+      "loss": 1.5167,
+      "step": 5976
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.417366773828636e-05,
+      "loss": 1.672,
+      "step": 5977
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.4054163098310365e-05,
+      "loss": 3.0026,
+      "step": 5978
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.3934774644510304e-05,
+      "loss": 0.3629,
+      "step": 5979
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.381550262482756e-05,
+      "loss": 1.6307,
+      "step": 5980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.3696347286961724e-05,
+      "loss": 3.8378,
+      "step": 5981
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.3577308878370024e-05,
+      "loss": 2.8168,
+      "step": 5982
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.345838764626688e-05,
+      "loss": 2.8246,
+      "step": 5983
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.333958383762339e-05,
+      "loss": 3.6833,
+      "step": 5984
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.322089769916686e-05,
+      "loss": 4.3263,
+      "step": 5985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.3102329477379876e-05,
+      "loss": 2.8645,
+      "step": 5986
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.298387941850052e-05,
+      "loss": 3.3965,
+      "step": 5987
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.2865547768521355e-05,
+      "loss": 2.5483,
+      "step": 5988
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.2747334773189004e-05,
+      "loss": 2.597,
+      "step": 5989
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.262924067800373e-05,
+      "loss": 3.8039,
+      "step": 5990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.251126572821882e-05,
+      "loss": 1.6216,
+      "step": 5991
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.239341016884014e-05,
+      "loss": 1.0965,
+      "step": 5992
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.227567424462563e-05,
+      "loss": 2.0156,
+      "step": 5993
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.2158058200084724e-05,
+      "loss": 1.4756,
+      "step": 5994
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.204056227947797e-05,
+      "loss": 2.9758,
+      "step": 5995
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.192318672681637e-05,
+      "loss": 2.9157,
+      "step": 5996
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.1805931785861e-05,
+      "loss": 2.0344,
+      "step": 5997
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.1688797700122426e-05,
+      "loss": 2.6333,
+      "step": 5998
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.1571784712860276e-05,
+      "loss": 3.8556,
+      "step": 5999
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.145489306708261e-05,
+      "loss": 4.3464,
+      "step": 6000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.133812300554556e-05,
+      "loss": 3.6759,
+      "step": 6001
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.122147477075271e-05,
+      "loss": 3.1033,
+      "step": 6002
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.1104948604954676e-05,
+      "loss": 4.4292,
+      "step": 6003
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.0988544750148515e-05,
+      "loss": 1.2505,
+      "step": 6004
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.0872263448077316e-05,
+      "loss": 1.8809,
+      "step": 6005
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.0756104940229664e-05,
+      "loss": 5.5241,
+      "step": 6006
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.0640069467839085e-05,
+      "loss": 3.6324,
+      "step": 6007
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.052415727188361e-05,
+      "loss": 3.0472,
+      "step": 6008
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.040836859308528e-05,
+      "loss": 4.1425,
+      "step": 6009
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.029270367190958e-05,
+      "loss": 2.4953,
+      "step": 6010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.017716274856501e-05,
+      "loss": 2.1265,
+      "step": 6011
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.0061746063002536e-05,
+      "loss": 4.1268,
+      "step": 6012
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.0061746063002536e-05,
+      "loss": 2.71,
+      "step": 6013
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.994645385491512e-05,
+      "loss": 3.6848,
+      "step": 6014
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.983128636373722e-05,
+      "loss": 2.5122,
+      "step": 6015
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9716243828644264e-05,
+      "loss": 3.0973,
+      "step": 6016
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9601326488552206e-05,
+      "loss": 1.587,
+      "step": 6017
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.948653458211713e-05,
+      "loss": 3.1695,
+      "step": 6018
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9371868347734185e-05,
+      "loss": 2.8316,
+      "step": 6019
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.925732802353797e-05,
+      "loss": 3.0949,
+      "step": 6020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.914291384740148e-05,
+      "loss": 3.1059,
+      "step": 6021
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9028626056935656e-05,
+      "loss": 4.7246,
+      "step": 6022
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.891446488948905e-05,
+      "loss": 5.3046,
+      "step": 6023
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.8800430582147175e-05,
+      "loss": 5.4482,
+      "step": 6024
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.8686523371732165e-05,
+      "loss": 4.2993,
+      "step": 6025
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.857274349480213e-05,
+      "loss": 5.0104,
+      "step": 6026
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.8459091187650786e-05,
+      "loss": 2.2338,
+      "step": 6027
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.83455666863069e-05,
+      "loss": 2.3386,
+      "step": 6028
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.8232170226533806e-05,
+      "loss": 0.8068,
+      "step": 6029
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.811890204382895e-05,
+      "loss": 2.4326,
+      "step": 6030
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.800576237342336e-05,
+      "loss": 5.0362,
+      "step": 6031
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.789275145028118e-05,
+      "loss": 4.368,
+      "step": 6032
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.777986950909919e-05,
+      "loss": 2.8653,
+      "step": 6033
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.7667116784306255e-05,
+      "loss": 3.963,
+      "step": 6034
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.755449351006296e-05,
+      "loss": 3.7769,
+      "step": 6035
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.744199992026101e-05,
+      "loss": 3.848,
+      "step": 6036
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.732963624852277e-05,
+      "loss": 3.3243,
+      "step": 6037
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.7217402728200844e-05,
+      "loss": 4.5874,
+      "step": 6038
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.7105299592377515e-05,
+      "loss": 3.9071,
+      "step": 6039
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.699332707386428e-05,
+      "loss": 2.6786,
+      "step": 6040
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.68814854052014e-05,
+      "loss": 2.4693,
+      "step": 6041
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.676977481865737e-05,
+      "loss": 1.8431,
+      "step": 6042
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.6658195546228455e-05,
+      "loss": 4.7941,
+      "step": 6043
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.654674781963824e-05,
+      "loss": 3.3933,
+      "step": 6044
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.643543187033711e-05,
+      "loss": 1.1196,
+      "step": 6045
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.632424792950175e-05,
+      "loss": 3.1615,
+      "step": 6046
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.621319622803474e-05,
+      "loss": 2.7866,
+      "step": 6047
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.610227699656401e-05,
+      "loss": 1.309,
+      "step": 6048
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.59914904654425e-05,
+      "loss": 1.8432,
+      "step": 6049
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.5880836864747193e-05,
+      "loss": 3.5613,
+      "step": 6050
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.577031642427937e-05,
+      "loss": 2.4352,
+      "step": 6051
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.5659929373563674e-05,
+      "loss": 2.4416,
+      "step": 6052
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.55496759418477e-05,
+      "loss": 3.3489,
+      "step": 6053
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.543955635810159e-05,
+      "loss": 3.9892,
+      "step": 6054
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.532957085101746e-05,
+      "loss": 3.1607,
+      "step": 6055
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.521971964900901e-05,
+      "loss": 2.7174,
+      "step": 6056
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.511000298021103e-05,
+      "loss": 4.4516,
+      "step": 6057
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.50004210724789e-05,
+      "loss": 3.1547,
+      "step": 6058
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.489097415338812e-05,
+      "loss": 5.4284,
+      "step": 6059
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.47816624502339e-05,
+      "loss": 4.4318,
+      "step": 6060
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.4672486190030594e-05,
+      "loss": 2.6713,
+      "step": 6061
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.456344559951128e-05,
+      "loss": 4.518,
+      "step": 6062
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.445454090512729e-05,
+      "loss": 0.9691,
+      "step": 6063
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.4345772333047747e-05,
+      "loss": 2.8891,
+      "step": 6064
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.423714010915904e-05,
+      "loss": 1.3063,
+      "step": 6065
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.412864445906444e-05,
+      "loss": 4.3212,
+      "step": 6066
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.402028560808356e-05,
+      "loss": 2.9813,
+      "step": 6067
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.391206378125191e-05,
+      "loss": 2.7975,
+      "step": 6068
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.380397920332046e-05,
+      "loss": 5.8986,
+      "step": 6069
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.369603209875513e-05,
+      "loss": 3.6916,
+      "step": 6070
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3588222691736326e-05,
+      "loss": 3.8732,
+      "step": 6071
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3480551206158514e-05,
+      "loss": 3.392,
+      "step": 6072
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.337301786562973e-05,
+      "loss": 3.4131,
+      "step": 6073
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.326562289347109e-05,
+      "loss": 2.2479,
+      "step": 6074
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.315836651271638e-05,
+      "loss": 4.0852,
+      "step": 6075
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.3051248946111534e-05,
+      "loss": 3.909,
+      "step": 6076
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.294427041611424e-05,
+      "loss": 2.4097,
+      "step": 6077
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.283743114489342e-05,
+      "loss": 1.3252,
+      "step": 6078
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.273073135432877e-05,
+      "loss": 2.232,
+      "step": 6079
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.2624171266010364e-05,
+      "loss": 4.8427,
+      "step": 6080
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.2517751101238226e-05,
+      "loss": 5.771,
+      "step": 6081
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.241147108102144e-05,
+      "loss": 2.4788,
+      "step": 6082
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.2305331426078425e-05,
+      "loss": 2.8383,
+      "step": 6083
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.219933235683594e-05,
+      "loss": 2.3156,
+      "step": 6084
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.209347409342882e-05,
+      "loss": 2.4718,
+      "step": 6085
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.1987756855699455e-05,
+      "loss": 1.7443,
+      "step": 6086
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.188218086319736e-05,
+      "loss": 1.384,
+      "step": 6087
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.177674633517873e-05,
+      "loss": 3.9336,
+      "step": 6088
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.1671453490605964e-05,
+      "loss": 3.7419,
+      "step": 6089
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.1566302548147196e-05,
+      "loss": 1.0984,
+      "step": 6090
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.146129372617591e-05,
+      "loss": 2.9744,
+      "step": 6091
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.13564272427704e-05,
+      "loss": 3.9736,
+      "step": 6092
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.125170331571339e-05,
+      "loss": 3.9049,
+      "step": 6093
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.1147122162491506e-05,
+      "loss": 2.1742,
+      "step": 6094
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.10426840002949e-05,
+      "loss": 2.2516,
+      "step": 6095
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.0938389046016744e-05,
+      "loss": 3.1198,
+      "step": 6096
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.083423751625284e-05,
+      "loss": 2.458,
+      "step": 6097
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.073022962730108e-05,
+      "loss": 3.7552,
+      "step": 6098
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.062636559516108e-05,
+      "loss": 1.9468,
+      "step": 6099
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.05226456355337e-05,
+      "loss": 2.6,
+      "step": 6100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.041906996382058e-05,
+      "loss": 4.1575,
+      "step": 6101
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.0315638795123714e-05,
+      "loss": 3.5562,
+      "step": 6102
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.021235234424501e-05,
+      "loss": 2.9332,
+      "step": 6103
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.010921082568582e-05,
+      "loss": 3.5408,
+      "step": 6104
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.00062144536465e-05,
+      "loss": 0.9623,
+      "step": 6105
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9903363442025978e-05,
+      "loss": 1.3826,
+      "step": 6106
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.980065800442132e-05,
+      "loss": 4.7299,
+      "step": 6107
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9698098354127234e-05,
+      "loss": 3.6079,
+      "step": 6108
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9595684704135706e-05,
+      "loss": 3.9722,
+      "step": 6109
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.949341726713546e-05,
+      "loss": 2.1828,
+      "step": 6110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9391296255511625e-05,
+      "loss": 2.1668,
+      "step": 6111
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9289321881345332e-05,
+      "loss": 2.2901,
+      "step": 6112
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9187494356412814e-05,
+      "loss": 3.7254,
+      "step": 6113
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.908581389218571e-05,
+      "loss": 3.9085,
+      "step": 6114
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.8984280699830112e-05,
+      "loss": 2.7981,
+      "step": 6115
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.8882894990206287e-05,
+      "loss": 0.91,
+      "step": 6116
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.87816569738682e-05,
+      "loss": 2.0731,
+      "step": 6117
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.8680566861063084e-05,
+      "loss": 2.0979,
+      "step": 6118
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.8579624861731035e-05,
+      "loss": 3.6826,
+      "step": 6119
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.847883118550453e-05,
+      "loss": 4.8974,
+      "step": 6120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.837818604170802e-05,
+      "loss": 3.4236,
+      "step": 6121
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.8277689639357506e-05,
+      "loss": 5.1749,
+      "step": 6122
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.817734218716007e-05,
+      "loss": 1.6593,
+      "step": 6123
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.8077143893513457e-05,
+      "loss": 3.5089,
+      "step": 6124
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7977094966505668e-05,
+      "loss": 3.3129,
+      "step": 6125
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7877195613914485e-05,
+      "loss": 3.4752,
+      "step": 6126
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7777446043207068e-05,
+      "loss": 3.9744,
+      "step": 6127
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7677846461539515e-05,
+      "loss": 1.2724,
+      "step": 6128
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7578397075756413e-05,
+      "loss": 3.5963,
+      "step": 6129
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7479098092390465e-05,
+      "loss": 2.8371,
+      "step": 6130
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7379949717662003e-05,
+      "loss": 2.952,
+      "step": 6131
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7280952157478577e-05,
+      "loss": 2.5856,
+      "step": 6132
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.718210561743453e-05,
+      "loss": 4.0481,
+      "step": 6133
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.7083410302810595e-05,
+      "loss": 3.7598,
+      "step": 6134
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.6984866418573416e-05,
+      "loss": 3.2228,
+      "step": 6135
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.6886474169375174e-05,
+      "loss": 3.3073,
+      "step": 6136
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.6788233759553115e-05,
+      "loss": 2.7584,
+      "step": 6137
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.6690145393129194e-05,
+      "loss": 1.3088,
+      "step": 6138
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.6592209273809564e-05,
+      "loss": 3.2059,
+      "step": 6139
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.649442560498423e-05,
+      "loss": 4.7977,
+      "step": 6140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.639679458972655e-05,
+      "loss": 3.6739,
+      "step": 6141
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.6299316430792908e-05,
+      "loss": 4.9994,
+      "step": 6142
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.620199133062222e-05,
+      "loss": 3.5982,
+      "step": 6143
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.610481949133564e-05,
+      "loss": 1.9648,
+      "step": 6144
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.600780111473571e-05,
+      "loss": 3.9838,
+      "step": 6145
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.5910936402306584e-05,
+      "loss": 3.5009,
+      "step": 6146
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.58142255552132e-05,
+      "loss": 1.8191,
+      "step": 6147
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.5717668774300974e-05,
+      "loss": 2.5185,
+      "step": 6148
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.56212662600953e-05,
+      "loss": 3.0334,
+      "step": 6149
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.5525018212801266e-05,
+      "loss": 2.0125,
+      "step": 6150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.5428924832303124e-05,
+      "loss": 3.3181,
+      "step": 6151
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.5332986318163933e-05,
+      "loss": 3.7138,
+      "step": 6152
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.5237202869625144e-05,
+      "loss": 3.4425,
+      "step": 6153
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.5141574685606152e-05,
+      "loss": 4.1948,
+      "step": 6154
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.5046101964703916e-05,
+      "loss": 1.6911,
+      "step": 6155
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.4950784905192524e-05,
+      "loss": 2.8704,
+      "step": 6156
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.4855623705022835e-05,
+      "loss": 4.3117,
+      "step": 6157
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.476061856182196e-05,
+      "loss": 3.7995,
+      "step": 6158
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.4665769672892957e-05,
+      "loss": 2.5638,
+      "step": 6159
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.457107723521439e-05,
+      "loss": 0.8058,
+      "step": 6160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.4476541445439872e-05,
+      "loss": 1.5045,
+      "step": 6161
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.4382162499897732e-05,
+      "loss": 2.9743,
+      "step": 6162
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.4287940594590554e-05,
+      "loss": 1.767,
+      "step": 6163
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.4193875925194787e-05,
+      "loss": 2.9225,
+      "step": 6164
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.409996868706035e-05,
+      "loss": 4.0139,
+      "step": 6165
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.400621907521019e-05,
+      "loss": 1.9624,
+      "step": 6166
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.391262728433994e-05,
+      "loss": 4.3364,
+      "step": 6167
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.381919350881744e-05,
+      "loss": 2.8636,
+      "step": 6168
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3725917942682373e-05,
+      "loss": 2.7336,
+      "step": 6169
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.36328007796459e-05,
+      "loss": 1.7075,
+      "step": 6170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.353984221309017e-05,
+      "loss": 3.7708,
+      "step": 6171
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3447042436067966e-05,
+      "loss": 3.5457,
+      "step": 6172
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3354401641302347e-05,
+      "loss": 2.9583,
+      "step": 6173
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3261920021186157e-05,
+      "loss": 2.3581,
+      "step": 6174
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3169597767781703e-05,
+      "loss": 2.7314,
+      "step": 6175
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.3077435072820408e-05,
+      "loss": 4.7385,
+      "step": 6176
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2985432127702024e-05,
+      "loss": 3.8741,
+      "step": 6177
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2893589123494853e-05,
+      "loss": 3.1074,
+      "step": 6178
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2801906250934946e-05,
+      "loss": 1.017,
+      "step": 6179
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2710383700425787e-05,
+      "loss": 4.9479,
+      "step": 6180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2619021662037898e-05,
+      "loss": 1.9688,
+      "step": 6181
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2527820325508486e-05,
+      "loss": 2.725,
+      "step": 6182
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2436779880240987e-05,
+      "loss": 1.7977,
+      "step": 6183
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.234590051530474e-05,
+      "loss": 3.2881,
+      "step": 6184
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2255182419434507e-05,
+      "loss": 2.942,
+      "step": 6185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.216462578103019e-05,
+      "loss": 2.6304,
+      "step": 6186
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.2074230788156336e-05,
+      "loss": 4.6406,
+      "step": 6187
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.198399762854182e-05,
+      "loss": 2.2502,
+      "step": 6188
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.189392648957942e-05,
+      "loss": 2.0339,
+      "step": 6189
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.180401755832543e-05,
+      "loss": 2.0202,
+      "step": 6190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.171427102149928e-05,
+      "loss": 1.6426,
+      "step": 6191
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.1624687065483163e-05,
+      "loss": 4.0984,
+      "step": 6192
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.1535265876321596e-05,
+      "loss": 5.9903,
+      "step": 6193
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.1446007639721098e-05,
+      "loss": 4.0182,
+      "step": 6194
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.1356912541049757e-05,
+      "loss": 4.1813,
+      "step": 6195
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.1267980765336883e-05,
+      "loss": 3.1522,
+      "step": 6196
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.117921249727257e-05,
+      "loss": 3.744,
+      "step": 6197
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.1090607921207385e-05,
+      "loss": 3.3721,
+      "step": 6198
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.1002167221151903e-05,
+      "loss": 1.9707,
+      "step": 6199
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.0913890580776406e-05,
+      "loss": 1.4835,
+      "step": 6200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.082577818341044e-05,
+      "loss": 1.9259,
+      "step": 6201
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.0737830212042476e-05,
+      "loss": 2.3918,
+      "step": 6202
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.0650046849319482e-05,
+      "loss": 1.3536,
+      "step": 6203
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.0562428277546596e-05,
+      "loss": 3.4266,
+      "step": 6204
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.047497467868672e-05,
+      "loss": 2.0818,
+      "step": 6205
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.038768623436016e-05,
+      "loss": 3.743,
+      "step": 6206
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.0300563125844307e-05,
+      "loss": 2.3505,
+      "step": 6207
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.02136055340729e-05,
+      "loss": 1.8056,
+      "step": 6208
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.0126813639636257e-05,
+      "loss": 3.2278,
+      "step": 6209
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.004018762278044e-05,
+      "loss": 2.191,
+      "step": 6210
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9953727663407072e-05,
+      "loss": 4.0857,
+      "step": 6211
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.98674339410729e-05,
+      "loss": 2.6188,
+      "step": 6212
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9781306634989406e-05,
+      "loss": 5.0469,
+      "step": 6213
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9695345924022514e-05,
+      "loss": 3.6253,
+      "step": 6214
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.960955198669211e-05,
+      "loss": 2.7627,
+      "step": 6215
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9523925001171785e-05,
+      "loss": 3.0192,
+      "step": 6216
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9438465145288398e-05,
+      "loss": 4.2517,
+      "step": 6217
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9353172596521686e-05,
+      "loss": 1.5633,
+      "step": 6218
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9268047532003987e-05,
+      "loss": 2.3648,
+      "step": 6219
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9183090128519742e-05,
+      "loss": 4.0676,
+      "step": 6220
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.909830056250528e-05,
+      "loss": 3.9377,
+      "step": 6221
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.90136790100483e-05,
+      "loss": 1.3158,
+      "step": 6222
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8929225646887627e-05,
+      "loss": 5.5183,
+      "step": 6223
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.884494064841278e-05,
+      "loss": 4.1386,
+      "step": 6224
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8760824189663607e-05,
+      "loss": 1.7996,
+      "step": 6225
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.867687644532996e-05,
+      "loss": 1.1601,
+      "step": 6226
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8593097589751306e-05,
+      "loss": 3.3157,
+      "step": 6227
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.850948779691638e-05,
+      "loss": 3.9649,
+      "step": 6228
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8426047240462796e-05,
+      "loss": 3.2738,
+      "step": 6229
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8342776093676705e-05,
+      "loss": 2.7788,
+      "step": 6230
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8259674529492465e-05,
+      "loss": 1.9224,
+      "step": 6231
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.81767427204922e-05,
+      "loss": 4.6436,
+      "step": 6232
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8093980838905543e-05,
+      "loss": 2.4693,
+      "step": 6233
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.8011389056609206e-05,
+      "loss": 2.0729,
+      "step": 6234
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.792896754512663e-05,
+      "loss": 2.3614,
+      "step": 6235
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.784671647562768e-05,
+      "loss": 3.9685,
+      "step": 6236
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7764636018928216e-05,
+      "loss": 2.2275,
+      "step": 6237
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7682726345489798e-05,
+      "loss": 2.7879,
+      "step": 6238
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.760098762541942e-05,
+      "loss": 2.8719,
+      "step": 6239
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7519420028468715e-05,
+      "loss": 2.5514,
+      "step": 6240
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7438023724034247e-05,
+      "loss": 2.2722,
+      "step": 6241
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7356798881156765e-05,
+      "loss": 1.9731,
+      "step": 6242
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7275745668520914e-05,
+      "loss": 2.9196,
+      "step": 6243
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7194864254454913e-05,
+      "loss": 2.5576,
+      "step": 6244
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7114154806930183e-05,
+      "loss": 3.8965,
+      "step": 6245
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7033617493561015e-05,
+      "loss": 2.0324,
+      "step": 6246
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6953252481604243e-05,
+      "loss": 4.5263,
+      "step": 6247
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6873059937958835e-05,
+      "loss": 1.5975,
+      "step": 6248
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6793040029165607e-05,
+      "loss": 3.3603,
+      "step": 6249
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.671319292140685e-05,
+      "loss": 1.3319,
+      "step": 6250
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6633518780505975e-05,
+      "loss": 2.2586,
+      "step": 6251
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.655401777192721e-05,
+      "loss": 3.4528,
+      "step": 6252
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.647469006077519e-05,
+      "loss": 2.6915,
+      "step": 6253
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6395535811794694e-05,
+      "loss": 4.7195,
+      "step": 6254
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.631655518937024e-05,
+      "loss": 1.9203,
+      "step": 6255
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.623774835752585e-05,
+      "loss": 3.8204,
+      "step": 6256
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6159115479924247e-05,
+      "loss": 2.4615,
+      "step": 6257
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.6080656719867494e-05,
+      "loss": 0.4674,
+      "step": 6258
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.600237224029564e-05,
+      "loss": 4.333,
+      "step": 6259
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.59242622037869e-05,
+      "loss": 1.8705,
+      "step": 6260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.584632677255722e-05,
+      "loss": 2.6193,
+      "step": 6261
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5768566108459903e-05,
+      "loss": 4.97,
+      "step": 6262
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5690980372985352e-05,
+      "loss": 0.3325,
+      "step": 6263
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5613569727260636e-05,
+      "loss": 4.6681,
+      "step": 6264
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.553633433204924e-05,
+      "loss": 3.3355,
+      "step": 6265
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5459274347750673e-05,
+      "loss": 2.013,
+      "step": 6266
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5382389934400164e-05,
+      "loss": 2.5964,
+      "step": 6267
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5305681251668335e-05,
+      "loss": 3.6838,
+      "step": 6268
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5229148458860832e-05,
+      "loss": 5.9373,
+      "step": 6269
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5152791714918136e-05,
+      "loss": 3.4782,
+      "step": 6270
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5076611178414835e-05,
+      "loss": 2.4857,
+      "step": 6271
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5000607007559831e-05,
+      "loss": 3.6606,
+      "step": 6272
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4924779360195695e-05,
+      "loss": 1.9151,
+      "step": 6273
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4849128393798383e-05,
+      "loss": 1.8451,
+      "step": 6274
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4773654265476932e-05,
+      "loss": 1.8196,
+      "step": 6275
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4698357131973128e-05,
+      "loss": 4.0176,
+      "step": 6276
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4623237149661173e-05,
+      "loss": 4.2632,
+      "step": 6277
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4548294474547386e-05,
+      "loss": 0.8965,
+      "step": 6278
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4473529262269848e-05,
+      "loss": 2.7762,
+      "step": 6279
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4398941668098087e-05,
+      "loss": 1.9712,
+      "step": 6280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4324531846932777e-05,
+      "loss": 5.2242,
+      "step": 6281
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4250299953305379e-05,
+      "loss": 4.3132,
+      "step": 6282
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4176246141377868e-05,
+      "loss": 3.8149,
+      "step": 6283
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4102370564942358e-05,
+      "loss": 3.3633,
+      "step": 6284
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.4028673377420832e-05,
+      "loss": 4.8869,
+      "step": 6285
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3955154731864783e-05,
+      "loss": 2.8815,
+      "step": 6286
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3881814780954928e-05,
+      "loss": 0.4875,
+      "step": 6287
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3808653677000972e-05,
+      "loss": 2.9668,
+      "step": 6288
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3735671571940822e-05,
+      "loss": 3.8122,
+      "step": 6289
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3662868617341195e-05,
+      "loss": 2.9008,
+      "step": 6290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3590244964396392e-05,
+      "loss": 3.1235,
+      "step": 6291
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3517800763928445e-05,
+      "loss": 2.4204,
+      "step": 6292
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3445536166386707e-05,
+      "loss": 2.7074,
+      "step": 6293
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3373451321847518e-05,
+      "loss": 3.7278,
+      "step": 6294
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.330154638001393e-05,
+      "loss": 3.7807,
+      "step": 6295
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3229821490215388e-05,
+      "loss": 4.671,
+      "step": 6296
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.3158276801407399e-05,
+      "loss": 3.3795,
+      "step": 6297
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.308691246217123e-05,
+      "loss": 3.752,
+      "step": 6298
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.301572862071363e-05,
+      "loss": 2.176,
+      "step": 6299
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2944725424866454e-05,
+      "loss": 2.5889,
+      "step": 6300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2873903022086454e-05,
+      "loss": 4.4758,
+      "step": 6301
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.280326155945496e-05,
+      "loss": 2.2166,
+      "step": 6302
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2732801183677301e-05,
+      "loss": 2.9789,
+      "step": 6303
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2662522041082958e-05,
+      "loss": 5.7633,
+      "step": 6304
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2592424277624981e-05,
+      "loss": 4.9589,
+      "step": 6305
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2522508038879731e-05,
+      "loss": 3.9465,
+      "step": 6306
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2452773470046575e-05,
+      "loss": 5.4769,
+      "step": 6307
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2383220715947607e-05,
+      "loss": 4.1197,
+      "step": 6308
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2313849921027309e-05,
+      "loss": 1.8252,
+      "step": 6309
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2244661229352317e-05,
+      "loss": 2.2793,
+      "step": 6310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2175654784611058e-05,
+      "loss": 2.3112,
+      "step": 6311
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.2106830730113472e-05,
+      "loss": 5.2345,
+      "step": 6312
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.203818920879074e-05,
+      "loss": 0.9074,
+      "step": 6313
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.196973036319492e-05,
+      "loss": 1.7698,
+      "step": 6314
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1901454335498763e-05,
+      "loss": 4.0775,
+      "step": 6315
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1833361267495279e-05,
+      "loss": 3.9189,
+      "step": 6316
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1765451300597585e-05,
+      "loss": 4.1024,
+      "step": 6317
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1697724575838475e-05,
+      "loss": 3.3629,
+      "step": 6318
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1630181233870319e-05,
+      "loss": 2.5356,
+      "step": 6319
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1562821414964386e-05,
+      "loss": 2.2216,
+      "step": 6320
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1495645259010956e-05,
+      "loss": 3.8143,
+      "step": 6321
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1428652905519088e-05,
+      "loss": 2.9638,
+      "step": 6322
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1361844493615847e-05,
+      "loss": 4.2234,
+      "step": 6323
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1295220162046405e-05,
+      "loss": 3.0314,
+      "step": 6324
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1228780049173604e-05,
+      "loss": 3.0532,
+      "step": 6325
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.116252429297775e-05,
+      "loss": 3.112,
+      "step": 6326
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1096453031056254e-05,
+      "loss": 4.1331,
+      "step": 6327
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.1030566400623388e-05,
+      "loss": 4.0168,
+      "step": 6328
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.096486453851e-05,
+      "loss": 2.9707,
+      "step": 6329
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.0899347581163189e-05,
+      "loss": 2.6322,
+      "step": 6330
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.083401566464608e-05,
+      "loss": 2.8542,
+      "step": 6331
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.0768868924637498e-05,
+      "loss": 3.674,
+      "step": 6332
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0703907496431787e-05,
+      "loss": 1.9706,
+      "step": 6333
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0639131514938194e-05,
+      "loss": 3.4675,
+      "step": 6334
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.057454111468108e-05,
+      "loss": 5.2749,
+      "step": 6335
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0510136429799322e-05,
+      "loss": 4.1038,
+      "step": 6336
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0445917594046106e-05,
+      "loss": 0.6047,
+      "step": 6337
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.038188474078866e-05,
+      "loss": 2.8867,
+      "step": 6338
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0318038003007957e-05,
+      "loss": 1.8992,
+      "step": 6339
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0254377513298475e-05,
+      "loss": 2.1651,
+      "step": 6340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.019090340386788e-05,
+      "loss": 3.7672,
+      "step": 6341
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0127615806536772e-05,
+      "loss": 2.7728,
+      "step": 6342
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0064514852738415e-05,
+      "loss": 0.962,
+      "step": 6343
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.0001600673518474e-05,
+      "loss": 1.9177,
+      "step": 6344
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.938873399534698e-06,
+      "loss": 1.9173,
+      "step": 6345
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.876333161056684e-06,
+      "loss": 1.9134,
+      "step": 6346
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.813980087965623e-06,
+      "loss": 3.5297,
+      "step": 6347
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.751814309754003e-06,
+      "loss": 3.6482,
+      "step": 6348
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.689835955525317e-06,
+      "loss": 2.428,
+      "step": 6349
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.628045153993847e-06,
+      "loss": 1.294,
+      "step": 6350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.566442033484446e-06,
+      "loss": 2.0209,
+      "step": 6351
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.505026721931965e-06,
+      "loss": 5.6328,
+      "step": 6352
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.443799346881388e-06,
+      "loss": 1.6561,
+      "step": 6353
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.382760035487526e-06,
+      "loss": 2.6109,
+      "step": 6354
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.32190891451442e-06,
+      "loss": 2.8279,
+      "step": 6355
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.261246110335341e-06,
+      "loss": 3.1278,
+      "step": 6356
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.200771748932502e-06,
+      "loss": 4.8996,
+      "step": 6357
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.140485955896727e-06,
+      "loss": 2.5471,
+      "step": 6358
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.080388856427258e-06,
+      "loss": 2.8646,
+      "step": 6359
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.020480575331457e-06,
+      "loss": 2.9402,
+      "step": 6360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.960761237024539e-06,
+      "loss": 3.2335,
+      "step": 6361
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.90123096552935e-06,
+      "loss": 1.4607,
+      "step": 6362
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.841889884476095e-06,
+      "loss": 3.8528,
+      "step": 6363
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.782738117102041e-06,
+      "loss": 2.5943,
+      "step": 6364
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.723775786251409e-06,
+      "loss": 4.645,
+      "step": 6365
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.665003014374751e-06,
+      "loss": 3.3056,
+      "step": 6366
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.606419923529207e-06,
+      "loss": 2.6291,
+      "step": 6367
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.548026635377882e-06,
+      "loss": 1.7001,
+      "step": 6368
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.489823271189734e-06,
+      "loss": 3.5233,
+      "step": 6369
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.431809951839287e-06,
+      "loss": 2.6175,
+      "step": 6370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.373986797806387e-06,
+      "loss": 1.9751,
+      "step": 6371
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.316353929175946e-06,
+      "loss": 4.4204,
+      "step": 6372
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.25891146563771e-06,
+      "loss": 2.3964,
+      "step": 6373
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.201659526485972e-06,
+      "loss": 1.6989,
+      "step": 6374
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.144598230619393e-06,
+      "loss": 2.0023,
+      "step": 6375
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.08772769654068e-06,
+      "loss": 3.3652,
+      "step": 6376
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.031048042356392e-06,
+      "loss": 3.9343,
+      "step": 6377
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.97455938577667e-06,
+      "loss": 4.7846,
+      "step": 6378
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.918261844114983e-06,
+      "loss": 2.081,
+      "step": 6379
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.86215553428793e-06,
+      "loss": 5.1071,
+      "step": 6380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.806240572814937e-06,
+      "loss": 2.97,
+      "step": 6381
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.750517075818142e-06,
+      "loss": 3.5441,
+      "step": 6382
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.694985159021827e-06,
+      "loss": 3.0968,
+      "step": 6383
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.639644937752621e-06,
+      "loss": 0.4212,
+      "step": 6384
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.584496526938922e-06,
+      "loss": 4.7359,
+      "step": 6385
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.529540041110972e-06,
+      "loss": 1.0036,
+      "step": 6386
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.474775594400229e-06,
+      "loss": 4.2072,
+      "step": 6387
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.420203300539408e-06,
+      "loss": 0.8376,
+      "step": 6388
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.3658232728621715e-06,
+      "loss": 3.1404,
+      "step": 6389
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.31163562430287e-06,
+      "loss": 4.9918,
+      "step": 6390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.257640467396343e-06,
+      "loss": 1.526,
+      "step": 6391
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.203837914277667e-06,
+      "loss": 4.023,
+      "step": 6392
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.1502280766819305e-06,
+      "loss": 3.1979,
+      "step": 6393
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.096811065943998e-06,
+      "loss": 3.2078,
+      "step": 6394
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.043586992998286e-06,
+      "loss": 1.72,
+      "step": 6395
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.990555968378553e-06,
+      "loss": 1.6114,
+      "step": 6396
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.937718102217494e-06,
+      "loss": 2.7116,
+      "step": 6397
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.8850735042468415e-06,
+      "loss": 3.5613,
+      "step": 6398
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.832622283796863e-06,
+      "loss": 4.4339,
+      "step": 6399
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.780364549796259e-06,
+      "loss": 2.9481,
+      "step": 6400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.728300410771904e-06,
+      "loss": 3.1164,
+      "step": 6401
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.676429974848586e-06,
+      "loss": 3.0414,
+      "step": 6402
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.62475334974888e-06,
+      "loss": 2.6432,
+      "step": 6403
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.573270642792795e-06,
+      "loss": 2.5291,
+      "step": 6404
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.52198196089766e-06,
+      "loss": 3.6291,
+      "step": 6405
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.470887410577864e-06,
+      "loss": 3.0981,
+      "step": 6406
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.419987097944591e-06,
+      "loss": 3.4765,
+      "step": 6407
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.369281128705673e-06,
+      "loss": 3.8607,
+      "step": 6408
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.318769608165343e-06,
+      "loss": 3.2045,
+      "step": 6409
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.2684526412239875e-06,
+      "loss": 3.7687,
+      "step": 6410
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.218330332377953e-06,
+      "loss": 2.1634,
+      "step": 6411
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.1684027857193405e-06,
+      "loss": 2.8154,
+      "step": 6412
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.118670104935775e-06,
+      "loss": 1.7662,
+      "step": 6413
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.0691323933102395e-06,
+      "loss": 3.0041,
+      "step": 6414
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.0197897537206505e-06,
+      "loss": 2.0333,
+      "step": 6415
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.970642288639972e-06,
+      "loss": 3.3229,
+      "step": 6416
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.921690100135713e-06,
+      "loss": 2.1371,
+      "step": 6417
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.872933289870031e-06,
+      "loss": 3.3991,
+      "step": 6418
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.82437195909915e-06,
+      "loss": 3.8985,
+      "step": 6419
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.77600620867339e-06,
+      "loss": 3.0681,
+      "step": 6420
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.72783613903689e-06,
+      "loss": 3.8675,
+      "step": 6421
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.679861850227408e-06,
+      "loss": 1.6041,
+      "step": 6422
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.632083441876113e-06,
+      "loss": 2.4817,
+      "step": 6423
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.584501013207366e-06,
+      "loss": 3.8657,
+      "step": 6424
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.537114663038556e-06,
+      "loss": 2.0534,
+      "step": 6425
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.489924489779841e-06,
+      "loss": 2.9003,
+      "step": 6426
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.442930591433959e-06,
+      "loss": 4.7511,
+      "step": 6427
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.3961330655961095e-06,
+      "loss": 1.2632,
+      "step": 6428
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.349532009453484e-06,
+      "loss": 2.7423,
+      "step": 6429
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.303127519785434e-06,
+      "loss": 0.3287,
+      "step": 6430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.256919692963025e-06,
+      "loss": 1.9828,
+      "step": 6431
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.210908624948918e-06,
+      "loss": 0.3957,
+      "step": 6432
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.165094411297133e-06,
+      "loss": 2.0495,
+      "step": 6433
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.1194771471528734e-06,
+      "loss": 3.0548,
+      "step": 6434
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.074056927252347e-06,
+      "loss": 4.6119,
+      "step": 6435
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 5.028833845922509e-06,
+      "loss": 3.5825,
+      "step": 6436
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.983807997080947e-06,
+      "loss": 2.5699,
+      "step": 6437
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.938979474235605e-06,
+      "loss": 2.28,
+      "step": 6438
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.8943483704846585e-06,
+      "loss": 3.2007,
+      "step": 6439
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.849914778516262e-06,
+      "loss": 1.1429,
+      "step": 6440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.805678790608425e-06,
+      "loss": 4.6735,
+      "step": 6441
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.761640498628739e-06,
+      "loss": 2.9943,
+      "step": 6442
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.7177999940342554e-06,
+      "loss": 0.5822,
+      "step": 6443
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.6741573678712455e-06,
+      "loss": 4.3683,
+      "step": 6444
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.630712710775065e-06,
+      "loss": 2.3502,
+      "step": 6445
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.587466112969951e-06,
+      "loss": 1.7375,
+      "step": 6446
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.544417664268696e-06,
+      "loss": 3.8914,
+      "step": 6447
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.50156745407272e-06,
+      "loss": 4.9214,
+      "step": 6448
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.458915571371647e-06,
+      "loss": 0.8488,
+      "step": 6449
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.4164621047433796e-06,
+      "loss": 3.1668,
+      "step": 6450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.3742071423536145e-06,
+      "loss": 1.7417,
+      "step": 6451
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.332150771955823e-06,
+      "loss": 3.7385,
+      "step": 6452
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.2902930808910945e-06,
+      "loss": 3.237,
+      "step": 6453
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.24863415608786e-06,
+      "loss": 3.3828,
+      "step": 6454
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.207174084061782e-06,
+      "loss": 3.6178,
+      "step": 6455
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.165912950915574e-06,
+      "loss": 3.6246,
+      "step": 6456
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.124850842338757e-06,
+      "loss": 4.2075,
+      "step": 6457
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.083987843607562e-06,
+      "loss": 1.5398,
+      "step": 6458
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.0433240395847504e-06,
+      "loss": 4.4817,
+      "step": 6459
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.0028595147192594e-06,
+      "loss": 2.1743,
+      "step": 6460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.962594353046312e-06,
+      "loss": 3.5469,
+      "step": 6461
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.922528638187062e-06,
+      "loss": 3.7246,
+      "step": 6462
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.88266245334844e-06,
+      "loss": 1.9388,
+      "step": 6463
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8429958813230415e-06,
+      "loss": 4.6956,
+      "step": 6464
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8035290044888704e-06,
+      "loss": 0.6538,
+      "step": 6465
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.76426190480923e-06,
+      "loss": 4.5616,
+      "step": 6466
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.725194663832532e-06,
+      "loss": 0.7322,
+      "step": 6467
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.6863273626921323e-06,
+      "loss": 2.7729,
+      "step": 6468
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.6476600821061524e-06,
+      "loss": 1.9349,
+      "step": 6469
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.609192902377323e-06,
+      "loss": 1.4034,
+      "step": 6470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.5709259033928188e-06,
+      "loss": 3.4021,
+      "step": 6471
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.5328591646240805e-06,
+      "loss": 4.7079,
+      "step": 6472
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.494992765126648e-06,
+      "loss": 1.7579,
+      "step": 6473
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.457326783540016e-06,
+      "loss": 2.6891,
+      "step": 6474
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.4198612980874677e-06,
+      "loss": 2.9728,
+      "step": 6475
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.382596386575865e-06,
+      "loss": 2.6635,
+      "step": 6476
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.345532126395623e-06,
+      "loss": 1.3231,
+      "step": 6477
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.308668594520281e-06,
+      "loss": 2.8203,
+      "step": 6478
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2720058675066557e-06,
+      "loss": 5.8857,
+      "step": 6479
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.2355440214944967e-06,
+      "loss": 2.4834,
+      "step": 6480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.199283132206332e-06,
+      "loss": 3.14,
+      "step": 6481
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.16322327494748e-06,
+      "loss": 3.2803,
+      "step": 6482
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.1273645246056247e-06,
+      "loss": 3.9342,
+      "step": 6483
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.091706955650853e-06,
+      "loss": 2.87,
+      "step": 6484
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.0562506421354386e-06,
+      "loss": 3.7666,
+      "step": 6485
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.0209956576936926e-06,
+      "loss": 1.8912,
+      "step": 6486
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.9859420755418367e-06,
+      "loss": 2.294,
+      "step": 6487
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.9510899684777935e-06,
+      "loss": 2.2761,
+      "step": 6488
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.9164394088810997e-06,
+      "loss": 1.8593,
+      "step": 6489
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.881990468712703e-06,
+      "loss": 2.3959,
+      "step": 6490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.847743219514887e-06,
+      "loss": 3.0195,
+      "step": 6491
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.813697732410958e-06,
+      "loss": 3.7553,
+      "step": 6492
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.779854078105315e-06,
+      "loss": 2.5572,
+      "step": 6493
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.7462123268831687e-06,
+      "loss": 4.0104,
+      "step": 6494
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.7127725486104207e-06,
+      "loss": 2.7246,
+      "step": 6495
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.6795348127335305e-06,
+      "loss": 2.6244,
+      "step": 6496
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.6464991882793498e-06,
+      "loss": 2.2758,
+      "step": 6497
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.6136657438549984e-06,
+      "loss": 1.3675,
+      "step": 6498
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.5810345476477316e-06,
+      "loss": 1.3473,
+      "step": 6499
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.5486056674247527e-06,
+      "loss": 2.0926,
+      "step": 6500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.516379170533145e-06,
+      "loss": 2.7749,
+      "step": 6501
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.48435512389964e-06,
+      "loss": 3.8848,
+      "step": 6502
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.4525335940305706e-06,
+      "loss": 2.0724,
+      "step": 6503
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.420914647011674e-06,
+      "loss": 0.6665,
+      "step": 6504
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.389498348507957e-06,
+      "loss": 3.1602,
+      "step": 6505
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.3582847637636072e-06,
+      "loss": 0.9029,
+      "step": 6506
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.327273957601783e-06,
+      "loss": 3.6781,
+      "step": 6507
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.296465994424557e-06,
+      "loss": 3.699,
+      "step": 6508
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2658609382127715e-06,
+      "loss": 6.0432,
+      "step": 6509
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2354588525257405e-06,
+      "loss": 4.1937,
+      "step": 6510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.2052598005014136e-06,
+      "loss": 3.4138,
+      "step": 6511
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.1752638448560126e-06,
+      "loss": 2.8077,
+      "step": 6512
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.1454710478839625e-06,
+      "loss": 3.8175,
+      "step": 6513
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.115881471457881e-06,
+      "loss": 1.316,
+      "step": 6514
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.086495177028225e-06,
+      "loss": 2.982,
+      "step": 6515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.057312225623331e-06,
+      "loss": 1.771,
+      "step": 6516
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.028332677849243e-06,
+      "loss": 3.4188,
+      "step": 6517
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9995565938895733e-06,
+      "loss": 3.9421,
+      "step": 6518
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9709840335053854e-06,
+      "loss": 1.7964,
+      "step": 6519
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.942615056035102e-06,
+      "loss": 3.1631,
+      "step": 6520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9144497203943067e-06,
+      "loss": 2.5605,
+      "step": 6521
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8864880850757105e-06,
+      "loss": 2.7121,
+      "step": 6522
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8587302081489954e-06,
+      "loss": 0.7429,
+      "step": 6523
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8311761472605826e-06,
+      "loss": 1.5456,
+      "step": 6524
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8038259596337427e-06,
+      "loss": 3.9535,
+      "step": 6525
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7766797020682958e-06,
+      "loss": 1.2149,
+      "step": 6526
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7497374309405456e-06,
+      "loss": 2.5502,
+      "step": 6527
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7229992022031794e-06,
+      "loss": 3.8818,
+      "step": 6528
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6964650713851226e-06,
+      "loss": 0.8265,
+      "step": 6529
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6701350935914518e-06,
+      "loss": 3.9135,
+      "step": 6530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6440093235032484e-06,
+      "loss": 2.4209,
+      "step": 6531
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6180878153775113e-06,
+      "loss": 2.1705,
+      "step": 6532
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.592370623047057e-06,
+      "loss": 4.5423,
+      "step": 6533
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5668577999203516e-06,
+      "loss": 2.93,
+      "step": 6534
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5415493989814457e-06,
+      "loss": 1.7138,
+      "step": 6535
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5164454727898737e-06,
+      "loss": 0.9039,
+      "step": 6536
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4915460734805097e-06,
+      "loss": 0.6225,
+      "step": 6537
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4668512527634571e-06,
+      "loss": 2.3839,
+      "step": 6538
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4423610619240025e-06,
+      "loss": 1.685,
+      "step": 6539
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4180755518224397e-06,
+      "loss": 2.3188,
+      "step": 6540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.3939947728939474e-06,
+      "loss": 3.0313,
+      "step": 6541
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.3701187751485988e-06,
+      "loss": 3.9638,
+      "step": 6542
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.3464476081711308e-06,
+      "loss": 0.5306,
+      "step": 6543
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.322981321120953e-06,
+      "loss": 1.7167,
+      "step": 6544
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2997199627318934e-06,
+      "loss": 3.1943,
+      "step": 6545
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2766635813123207e-06,
+      "loss": 2.9239,
+      "step": 6546
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2538122247448213e-06,
+      "loss": 1.6211,
+      "step": 6547
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.231165940486223e-06,
+      "loss": 3.144,
+      "step": 6548
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2087247755674491e-06,
+      "loss": 2.1638,
+      "step": 6549
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.186488776593464e-06,
+      "loss": 4.8036,
+      "step": 6550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.1644579897431395e-06,
+      "loss": 1.9299,
+      "step": 6551
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.1426324607691664e-06,
+      "loss": 2.0585,
+      "step": 6552
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.1210122349979758e-06,
+      "loss": 4.1056,
+      "step": 6553
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.0995973573296513e-06,
+      "loss": 5.547,
+      "step": 6554
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.0783878722377405e-06,
+      "loss": 1.7206,
+      "step": 6555
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.057383823769309e-06,
+      "loss": 0.503,
+      "step": 6556
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.036585255544764e-06,
+      "loss": 4.0372,
+      "step": 6557
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.0159922107577879e-06,
+      "loss": 1.7859,
+      "step": 6558
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.95604732175226e-07,
+      "loss": 4.2338,
+      "step": 6559
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.754228621370098e-07,
+      "loss": 2.4224,
+      "step": 6560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.554466425560792e-07,
+      "loss": 2.9284,
+      "step": 6561
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.356761149182825e-07,
+      "loss": 4.182,
+      "step": 6562
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.161113202822868e-07,
+      "loss": 1.771,
+      "step": 6563
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.967522992795241e-07,
+      "loss": 1.4067,
+      "step": 6564
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.775990921140564e-07,
+      "loss": 2.0825,
+      "step": 6565
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.586517385625548e-07,
+      "loss": 3.275,
+      "step": 6566
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.399102779741319e-07,
+      "loss": 1.6155,
+      "step": 6567
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.213747492703538e-07,
+      "loss": 2.3379,
+      "step": 6568
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.030451909450842e-07,
+      "loss": 2.9097,
+      "step": 6569
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.849216410644622e-07,
+      "loss": 2.7449,
+      "step": 6570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.670041372667802e-07,
+      "loss": 4.98,
+      "step": 6571
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.492927167624508e-07,
+      "loss": 4.4808,
+      "step": 6572
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.317874163338178e-07,
+      "loss": 3.6924,
+      "step": 6573
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.144882723352564e-07,
+      "loss": 2.3378,
+      "step": 6574
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.973953206929617e-07,
+      "loss": 5.1302,
+      "step": 6575
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.805085969049274e-07,
+      "loss": 0.5724,
+      "step": 6576
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.638281360408228e-07,
+      "loss": 2.2996,
+      "step": 6577
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.473539727420486e-07,
+      "loss": 5.1154,
+      "step": 6578
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.310861412214819e-07,
+      "loss": 2.9058,
+      "step": 6579
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.150246752635535e-07,
+      "loss": 3.484,
+      "step": 6580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.991696082240816e-07,
+      "loss": 3.6189,
+      "step": 6581
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.835209730302715e-07,
+      "loss": 2.5889,
+      "step": 6582
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.68078802180605e-07,
+      "loss": 3.7001,
+      "step": 6583
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.528431277448065e-07,
+      "loss": 0.5279,
+      "step": 6584
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.378139813637328e-07,
+      "loss": 2.2988,
+      "step": 6585
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.229913942493613e-07,
+      "loss": 2.1892,
+      "step": 6586
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.08375397184635e-07,
+      "loss": 1.7133,
+      "step": 6587
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.939660205235174e-07,
+      "loss": 2.3571,
+      "step": 6588
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.797632941908492e-07,
+      "loss": 3.316,
+      "step": 6589
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.657672476823138e-07,
+      "loss": 3.6592,
+      "step": 6590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.5197791006437174e-07,
+      "loss": 1.8347,
+      "step": 6591
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.383953099741711e-07,
+      "loss": 2.1355,
+      "step": 6592
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.250194756195591e-07,
+      "loss": 5.0032,
+      "step": 6593
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.118504347789487e-07,
+      "loss": 4.0111,
+      "step": 6594
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.9888821480130733e-07,
+      "loss": 2.7856,
+      "step": 6595
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8613284260605733e-07,
+      "loss": 2.0422,
+      "step": 6596
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7358434468308667e-07,
+      "loss": 1.016,
+      "step": 6597
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.612427470926383e-07,
+      "loss": 3.0032,
+      "step": 6598
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.4910807546525427e-07,
+      "loss": 1.6994,
+      "step": 6599
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.3718035500177604e-07,
+      "loss": 4.3098,
+      "step": 6600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.2545961047322213e-07,
+      "loss": 1.7938,
+      "step": 6601
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.1394586622078836e-07,
+      "loss": 1.3368,
+      "step": 6602
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.026391461558031e-07,
+      "loss": 4.2103,
+      "step": 6603
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.9153947375959445e-07,
+      "loss": 2.2568,
+      "step": 6604
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8064687208355646e-07,
+      "loss": 3.3037,
+      "step": 6605
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.6996136374902725e-07,
+      "loss": 3.634,
+      "step": 6606
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.594829709472557e-07,
+      "loss": 1.9874,
+      "step": 6607
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.4921171543937914e-07,
+      "loss": 3.3977,
+      "step": 6608
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.391476185563235e-07,
+      "loss": 3.9416,
+      "step": 6609
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.2929070119884767e-07,
+      "loss": 1.4704,
+      "step": 6610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.1964098383742137e-07,
+      "loss": 3.3239,
+      "step": 6611
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.1019848651219199e-07,
+      "loss": 2.1163,
+      "step": 6612
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.0096322883298435e-07,
+      "loss": 0.6761,
+      "step": 6613
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.919352299792232e-07,
+      "loss": 3.2203,
+      "step": 6614
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8311450869993307e-07,
+      "loss": 3.0551,
+      "step": 6615
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7450108331361624e-07,
+      "loss": 3.2846,
+      "step": 6616
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6609497170834155e-07,
+      "loss": 2.4274,
+      "step": 6617
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5789619134157774e-07,
+      "loss": 3.4998,
+      "step": 6618
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4990475924023807e-07,
+      "loss": 1.4307,
+      "step": 6619
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4212069200063573e-07,
+      "loss": 3.7367,
+      "step": 6620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.345440057884062e-07,
+      "loss": 2.6607,
+      "step": 6621
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2717471633854062e-07,
+      "loss": 2.459,
+      "step": 6622
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.200128389552857e-07,
+      "loss": 2.8668,
+      "step": 6623
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.13058388512155e-07,
+      "loss": 2.773,
+      "step": 6624
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.0631137945189551e-07,
+      "loss": 0.7701,
+      "step": 6625
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.977182578643219e-08,
+      "loss": 1.488,
+      "step": 6626
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.343974109685682e-08,
+      "loss": 2.5888,
+      "step": 6627
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.731513853340589e-08,
+      "loss": 4.4403,
+      "step": 6628
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.139803081540498e-08,
+      "loss": 2.4301,
+      "step": 6629
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.568843023130213e-08,
+      "loss": 5.436,
+      "step": 6630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.018634863856788e-08,
+      "loss": 3.2294,
+      "step": 6631
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.489179746370644e-08,
+      "loss": 1.3442,
+      "step": 6632
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.980478770224452e-08,
+      "loss": 1.231,
+      "step": 6633
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.492532991868693e-08,
+      "loss": 2.2293,
+      "step": 6634
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.025343424651663e-08,
+      "loss": 2.7612,
+      "step": 6635
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.578911038813916e-08,
+      "loss": 4.5106,
+      "step": 6636
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.1532367614893766e-08,
+      "loss": 1.2227,
+      "step": 6637
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7483214767009e-08,
+      "loss": 2.3901,
+      "step": 6638
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.364166025362492e-08,
+      "loss": 3.0601,
+      "step": 6639
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.0007712052737556e-08,
+      "loss": 2.061,
+      "step": 6640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.6581377711176747e-08,
+      "loss": 5.1304,
+      "step": 6641
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.3362664344639407e-08,
+      "loss": 0.8038,
+      "step": 6642
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.0351578637611833e-08,
+      "loss": 2.6071,
+      "step": 6643
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7548126843414115e-08,
+      "loss": 4.0471,
+      "step": 6644
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4952314784144607e-08,
+      "loss": 1.076,
+      "step": 6645
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2564147850679942e-08,
+      "loss": 6.6913,
+      "step": 6646
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.0383631002686134e-08,
+      "loss": 4.5755,
+      "step": 6647
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.41076876856306e-09,
+      "loss": 3.7389,
+      "step": 6648
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.64556524547777e-09,
+      "loss": 4.0228,
+      "step": 6649
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.088024099353383e-09,
+      "loss": 2.0675,
+      "step": 6650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.738148564813582e-09,
+      "loss": 4.0211,
+      "step": 6651
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.5959414452270124e-09,
+      "loss": 3.6333,
+      "step": 6652
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6614051127072928e-09,
+      "loss": 0.8735,
+      "step": 6653
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.345415080463936e-10,
+      "loss": 3.5458,
+      "step": 6654
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.1535214075905106e-10,
+      "loss": 3.3344,
+      "step": 6655
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.038380891049684e-10,
+      "loss": 1.0162,
+      "step": 6656
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0,
+      "loss": 3.6796,
+      "step": 6657
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.038380891049684e-10,
+      "loss": 2.7968,
+      "step": 6658
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.1535214075905106e-10,
+      "loss": 5.1606,
+      "step": 6659
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.345415080463936e-10,
+      "loss": 2.9525,
+      "step": 6660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6614051127072928e-09,
+      "loss": 1.7114,
+      "step": 6661
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.5959414452270124e-09,
+      "loss": 2.5069,
+      "step": 6662
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.738148564813582e-09,
+      "loss": 3.1014,
+      "step": 6663
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.088024099353383e-09,
+      "loss": 3.0959,
+      "step": 6664
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.64556524547777e-09,
+      "loss": 1.2974,
+      "step": 6665
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.41076876856306e-09,
+      "loss": 4.1083,
+      "step": 6666
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.0383631002686134e-08,
+      "loss": 2.7293,
+      "step": 6667
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.2564147850679942e-08,
+      "loss": 3.8114,
+      "step": 6668
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4952314784144607e-08,
+      "loss": 3.2248,
+      "step": 6669
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7548126843414115e-08,
+      "loss": 0.7698,
+      "step": 6670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.0351578637611833e-08,
+      "loss": 3.9617,
+      "step": 6671
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.3362664344639407e-08,
+      "loss": 3.9816,
+      "step": 6672
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.6581377711176747e-08,
+      "loss": 2.2305,
+      "step": 6673
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.0007712052737556e-08,
+      "loss": 3.1384,
+      "step": 6674
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.364166025362492e-08,
+      "loss": 2.2304,
+      "step": 6675
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7483214767009e-08,
+      "loss": 2.6186,
+      "step": 6676
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.153236761488266e-08,
+      "loss": 3.9991,
+      "step": 6677
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.578911038813916e-08,
+      "loss": 2.0531,
+      "step": 6678
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.025343424651663e-08,
+      "loss": 1.9292,
+      "step": 6679
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.492532991868693e-08,
+      "loss": 1.5134,
+      "step": 6680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.980478770223341e-08,
+      "loss": 1.1674,
+      "step": 6681
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.489179746369534e-08,
+      "loss": 1.9954,
+      "step": 6682
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.018634863855677e-08,
+      "loss": 1.0042,
+      "step": 6683
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.568843023130213e-08,
+      "loss": 3.5995,
+      "step": 6684
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.139803081540498e-08,
+      "loss": 1.3956,
+      "step": 6685
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.731513853339479e-08,
+      "loss": 2.7772,
+      "step": 6686
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.343974109684572e-08,
+      "loss": 2.619,
+      "step": 6687
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.977182578643219e-08,
+      "loss": 3.5951,
+      "step": 6688
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.0631137945189551e-07,
+      "loss": 1.4583,
+      "step": 6689
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.13058388512155e-07,
+      "loss": 2.6703,
+      "step": 6690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.200128389552857e-07,
+      "loss": 1.813,
+      "step": 6691
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.271747163385295e-07,
+      "loss": 2.7863,
+      "step": 6692
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.345440057884062e-07,
+      "loss": 2.4387,
+      "step": 6693
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.421206920006246e-07,
+      "loss": 4.1959,
+      "step": 6694
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4990475924023807e-07,
+      "loss": 1.9268,
+      "step": 6695
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.5789619134157774e-07,
+      "loss": 3.3108,
+      "step": 6696
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6609497170834155e-07,
+      "loss": 2.5215,
+      "step": 6697
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7450108331361624e-07,
+      "loss": 1.298,
+      "step": 6698
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.8311450869992198e-07,
+      "loss": 2.3697,
+      "step": 6699
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.919352299792232e-07,
+      "loss": 3.7122,
+      "step": 6700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.0096322883297324e-07,
+      "loss": 5.5776,
+      "step": 6701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.1019848651218087e-07,
+      "loss": 2.9944,
+      "step": 6702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.1964098383741027e-07,
+      "loss": 2.269,
+      "step": 6703
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2929070119884767e-07,
+      "loss": 0.7828,
+      "step": 6704
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.391476185563124e-07,
+      "loss": 1.2326,
+      "step": 6705
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.492117154393681e-07,
+      "loss": 2.7228,
+      "step": 6706
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.594829709472557e-07,
+      "loss": 3.1684,
+      "step": 6707
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.6996136374902725e-07,
+      "loss": 3.6527,
+      "step": 6708
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8064687208355646e-07,
+      "loss": 1.5937,
+      "step": 6709
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.9153947375959445e-07,
+      "loss": 1.5246,
+      "step": 6710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.026391461558031e-07,
+      "loss": 4.5758,
+      "step": 6711
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1394586622078836e-07,
+      "loss": 2.8364,
+      "step": 6712
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.2545961047321107e-07,
+      "loss": 3.4661,
+      "step": 6713
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.371803550017649e-07,
+      "loss": 4.2065,
+      "step": 6714
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.4910807546525427e-07,
+      "loss": 0.8699,
+      "step": 6715
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6124274709262717e-07,
+      "loss": 2.8485,
+      "step": 6716
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.7358434468308667e-07,
+      "loss": 2.263,
+      "step": 6717
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8613284260605733e-07,
+      "loss": 1.3417,
+      "step": 6718
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.988882148012962e-07,
+      "loss": 3.7401,
+      "step": 6719
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.118504347789487e-07,
+      "loss": 1.6783,
+      "step": 6720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.250194756195591e-07,
+      "loss": 2.885,
+      "step": 6721
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.383953099741711e-07,
+      "loss": 1.0245,
+      "step": 6722
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.519779100643606e-07,
+      "loss": 4.3143,
+      "step": 6723
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.657672476823138e-07,
+      "loss": 3.8703,
+      "step": 6724
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.797632941908492e-07,
+      "loss": 2.6768,
+      "step": 6725
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.939660205235064e-07,
+      "loss": 3.3152,
+      "step": 6726
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.083753971846239e-07,
+      "loss": 4.0794,
+      "step": 6727
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.229913942493503e-07,
+      "loss": 3.0877,
+      "step": 6728
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.378139813637217e-07,
+      "loss": 3.4126,
+      "step": 6729
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.528431277448065e-07,
+      "loss": 1.211,
+      "step": 6730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.68078802180605e-07,
+      "loss": 1.7563,
+      "step": 6731
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.835209730302604e-07,
+      "loss": 1.6993,
+      "step": 6732
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.991696082240705e-07,
+      "loss": 2.5246,
+      "step": 6733
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.150246752635424e-07,
+      "loss": 5.3541,
+      "step": 6734
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.310861412214819e-07,
+      "loss": 4.5902,
+      "step": 6735
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.473539727420374e-07,
+      "loss": 2.4475,
+      "step": 6736
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.638281360408228e-07,
+      "loss": 0.7215,
+      "step": 6737
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.805085969049164e-07,
+      "loss": 1.1261,
+      "step": 6738
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.973953206929506e-07,
+      "loss": 0.3838,
+      "step": 6739
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.144882723352453e-07,
+      "loss": 3.2145,
+      "step": 6740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.317874163338068e-07,
+      "loss": 1.8847,
+      "step": 6741
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.492927167624398e-07,
+      "loss": 1.9378,
+      "step": 6742
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.670041372667691e-07,
+      "loss": 3.6491,
+      "step": 6743
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.849216410644511e-07,
+      "loss": 6.3263,
+      "step": 6744
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.030451909450731e-07,
+      "loss": 1.7028,
+      "step": 6745
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.213747492703427e-07,
+      "loss": 1.0496,
+      "step": 6746
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.399102779741208e-07,
+      "loss": 1.1331,
+      "step": 6747
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.586517385625437e-07,
+      "loss": 2.6869,
+      "step": 6748
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.775990921140564e-07,
+      "loss": 3.2642,
+      "step": 6749
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.967522992795129e-07,
+      "loss": 1.0568,
+      "step": 6750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.161113202822757e-07,
+      "loss": 4.1171,
+      "step": 6751
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.356761149182713e-07,
+      "loss": 2.7979,
+      "step": 6752
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.554466425560682e-07,
+      "loss": 3.3152,
+      "step": 6753
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.754228621369988e-07,
+      "loss": 3.0276,
+      "step": 6754
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.95604732175215e-07,
+      "loss": 4.0269,
+      "step": 6755
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.0159922107577879e-06,
+      "loss": 4.1042,
+      "step": 6756
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.036585255544753e-06,
+      "loss": 2.1358,
+      "step": 6757
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.0573838237692979e-06,
+      "loss": 5.0422,
+      "step": 6758
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.0783878722377295e-06,
+      "loss": 1.9022,
+      "step": 6759
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.0995973573296403e-06,
+      "loss": 6.0838,
+      "step": 6760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.1210122349979645e-06,
+      "loss": 2.0764,
+      "step": 6761
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.1426324607691551e-06,
+      "loss": 4.1225,
+      "step": 6762
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.1644579897431285e-06,
+      "loss": 3.6839,
+      "step": 6763
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.186488776593453e-06,
+      "loss": 4.6974,
+      "step": 6764
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2087247755674381e-06,
+      "loss": 2.6616,
+      "step": 6765
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.231165940486212e-06,
+      "loss": 2.1955,
+      "step": 6766
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2538122247448103e-06,
+      "loss": 2.772,
+      "step": 6767
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2766635813123096e-06,
+      "loss": 3.1452,
+      "step": 6768
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.2997199627318824e-06,
+      "loss": 4.054,
+      "step": 6769
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3229813211209419e-06,
+      "loss": 2.5381,
+      "step": 6770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3464476081711198e-06,
+      "loss": 3.8155,
+      "step": 6771
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3701187751485878e-06,
+      "loss": 0.5761,
+      "step": 6772
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.3939947728939361e-06,
+      "loss": 2.3034,
+      "step": 6773
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4180755518224287e-06,
+      "loss": 2.8475,
+      "step": 6774
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4423610619239914e-06,
+      "loss": 3.1874,
+      "step": 6775
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466851252763446e-06,
+      "loss": 2.9013,
+      "step": 6776
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4915460734804987e-06,
+      "loss": 3.0727,
+      "step": 6777
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5164454727898624e-06,
+      "loss": 2.8424,
+      "step": 6778
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5415493989814344e-06,
+      "loss": 4.6959,
+      "step": 6779
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.5668577999203294e-06,
+      "loss": 3.4707,
+      "step": 6780
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.592370623047046e-06,
+      "loss": 3.3282,
+      "step": 6781
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6180878153775003e-06,
+      "loss": 1.8041,
+      "step": 6782
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6440093235032372e-06,
+      "loss": 2.8431,
+      "step": 6783
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6701350935914406e-06,
+      "loss": 1.8316,
+      "step": 6784
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6964650713851116e-06,
+      "loss": 3.9346,
+      "step": 6785
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7229992022031682e-06,
+      "loss": 1.7099,
+      "step": 6786
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7497374309405346e-06,
+      "loss": 3.4881,
+      "step": 6787
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7766797020682845e-06,
+      "loss": 3.8713,
+      "step": 6788
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8038259596337314e-06,
+      "loss": 4.1459,
+      "step": 6789
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8311761472605716e-06,
+      "loss": 2.0288,
+      "step": 6790
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8587302081489732e-06,
+      "loss": 3.297,
+      "step": 6791
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.8864880850756883e-06,
+      "loss": 1.65,
+      "step": 6792
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9144497203942957e-06,
+      "loss": 2.8155,
+      "step": 6793
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9426150560350797e-06,
+      "loss": 2.8181,
+      "step": 6794
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.970984033505374e-06,
+      "loss": 3.3081,
+      "step": 6795
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.999556593889551e-06,
+      "loss": 2.7892,
+      "step": 6796
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.028332677849232e-06,
+      "loss": 3.3983,
+      "step": 6797
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.05731222562332e-06,
+      "loss": 4.0381,
+      "step": 6798
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.086495177028203e-06,
+      "loss": 3.185,
+      "step": 6799
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.115881471457859e-06,
+      "loss": 2.8207,
+      "step": 6800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.1454710478839515e-06,
+      "loss": 3.8982,
+      "step": 6801
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.1752638448560016e-06,
+      "loss": 1.8658,
+      "step": 6802
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.205259800501391e-06,
+      "loss": 2.4203,
+      "step": 6803
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.235458852525729e-06,
+      "loss": 3.024,
+      "step": 6804
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.2658609382127494e-06,
+      "loss": 2.4082,
+      "step": 6805
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.296465994424546e-06,
+      "loss": 1.5478,
+      "step": 6806
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.327273957601772e-06,
+      "loss": 2.4656,
+      "step": 6807
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.358284763763585e-06,
+      "loss": 2.2351,
+      "step": 6808
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.389498348507946e-06,
+      "loss": 5.4562,
+      "step": 6809
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.4209146470116627e-06,
+      "loss": 2.5757,
+      "step": 6810
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.452533594030559e-06,
+      "loss": 4.9259,
+      "step": 6811
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.484355123899629e-06,
+      "loss": 1.8478,
+      "step": 6812
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.516379170533123e-06,
+      "loss": 1.3855,
+      "step": 6813
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.5486056674247417e-06,
+      "loss": 3.6289,
+      "step": 6814
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.581034547647709e-06,
+      "loss": 3.2495,
+      "step": 6815
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.613665743854987e-06,
+      "loss": 3.2851,
+      "step": 6816
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.6464991882793278e-06,
+      "loss": 3.5153,
+      "step": 6817
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.6795348127335085e-06,
+      "loss": 0.917,
+      "step": 6818
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.7127725486104093e-06,
+      "loss": 3.342,
+      "step": 6819
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.7462123268831573e-06,
+      "loss": 4.7729,
+      "step": 6820
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.779854078105304e-06,
+      "loss": 3.2708,
+      "step": 6821
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.813697732410947e-06,
+      "loss": 2.079,
+      "step": 6822
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.847743219514876e-06,
+      "loss": 3.151,
+      "step": 6823
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8819904687126806e-06,
+      "loss": 0.5074,
+      "step": 6824
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.9164394088810777e-06,
+      "loss": 2.7991,
+      "step": 6825
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.9510899684777715e-06,
+      "loss": 4.1094,
+      "step": 6826
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.9859420755418143e-06,
+      "loss": 3.1323,
+      "step": 6827
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.0209956576936816e-06,
+      "loss": 3.5263,
+      "step": 6828
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.0562506421354166e-06,
+      "loss": 2.709,
+      "step": 6829
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.0917069556508304e-06,
+      "loss": 2.7713,
+      "step": 6830
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1273645246056027e-06,
+      "loss": 3.5748,
+      "step": 6831
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.1632232749474576e-06,
+      "loss": 3.4853,
+      "step": 6832
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.19928313220631e-06,
+      "loss": 1.8093,
+      "step": 6833
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.2355440214944742e-06,
+      "loss": 2.6023,
+      "step": 6834
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.2720058675066332e-06,
+      "loss": 3.0735,
+      "step": 6835
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.308668594520259e-06,
+      "loss": 2.9663,
+      "step": 6836
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.3455321263956007e-06,
+      "loss": 3.1029,
+      "step": 6837
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.3825963865758535e-06,
+      "loss": 2.8486,
+      "step": 6838
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.4198612980874456e-06,
+      "loss": 0.823,
+      "step": 6839
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.457326783540005e-06,
+      "loss": 3.4472,
+      "step": 6840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.494992765126626e-06,
+      "loss": 1.4545,
+      "step": 6841
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.5328591646240585e-06,
+      "loss": 1.0023,
+      "step": 6842
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.5709259033927967e-06,
+      "loss": 2.1854,
+      "step": 6843
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.609192902377312e-06,
+      "loss": 4.0629,
+      "step": 6844
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6476600821061303e-06,
+      "loss": 4.1494,
+      "step": 6845
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6863273626921103e-06,
+      "loss": 1.1223,
+      "step": 6846
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.72519466383251e-06,
+      "loss": 1.4806,
+      "step": 6847
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.7642619048092074e-06,
+      "loss": 4.5071,
+      "step": 6848
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8035290044888484e-06,
+      "loss": 2.4503,
+      "step": 6849
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8429958813230195e-06,
+      "loss": 2.2924,
+      "step": 6850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.882662453348429e-06,
+      "loss": 4.3362,
+      "step": 6851
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.922528638187039e-06,
+      "loss": 1.899,
+      "step": 6852
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.96259435304629e-06,
+      "loss": 4.049,
+      "step": 6853
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.002859514719237e-06,
+      "loss": 4.1001,
+      "step": 6854
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.043324039584728e-06,
+      "loss": 2.4778,
+      "step": 6855
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.08398784360754e-06,
+      "loss": 2.5837,
+      "step": 6856
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.124850842338735e-06,
+      "loss": 2.9025,
+      "step": 6857
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.165912950915551e-06,
+      "loss": 1.188,
+      "step": 6858
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.20717408406176e-06,
+      "loss": 4.2652,
+      "step": 6859
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.248634156087838e-06,
+      "loss": 3.6222,
+      "step": 6860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.2902930808910724e-06,
+      "loss": 1.1694,
+      "step": 6861
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.3321507719558115e-06,
+      "loss": 2.7972,
+      "step": 6862
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.3742071423535925e-06,
+      "loss": 2.7552,
+      "step": 6863
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.4164621047433575e-06,
+      "loss": 5.1213,
+      "step": 6864
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.458915571371624e-06,
+      "loss": 1.6019,
+      "step": 6865
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.501567454072697e-06,
+      "loss": 1.2281,
+      "step": 6866
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.544417664268674e-06,
+      "loss": 4.3909,
+      "step": 6867
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.587466112969929e-06,
+      "loss": 1.9086,
+      "step": 6868
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.630712710775043e-06,
+      "loss": 2.3485,
+      "step": 6869
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.6741573678712235e-06,
+      "loss": 3.1735,
+      "step": 6870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.717799994034233e-06,
+      "loss": 2.6044,
+      "step": 6871
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.761640498628717e-06,
+      "loss": 2.79,
+      "step": 6872
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.805678790608403e-06,
+      "loss": 3.8526,
+      "step": 6873
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.84991477851624e-06,
+      "loss": 3.2279,
+      "step": 6874
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.894348370484636e-06,
+      "loss": 5.0025,
+      "step": 6875
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.938979474235583e-06,
+      "loss": 4.1778,
+      "step": 6876
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.983807997080925e-06,
+      "loss": 3.4596,
+      "step": 6877
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.028833845922487e-06,
+      "loss": 1.5599,
+      "step": 6878
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.074056927252324e-06,
+      "loss": 3.6133,
+      "step": 6879
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.119477147152851e-06,
+      "loss": 3.0812,
+      "step": 6880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.165094411297111e-06,
+      "loss": 1.8414,
+      "step": 6881
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5.210908624948896e-06,
+      "loss": 5.6403,
+      "step": 6882
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.256919692963003e-06,
+      "loss": 1.955,
+      "step": 6883
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.303127519785411e-06,
+      "loss": 2.8036,
+      "step": 6884
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.349532009453461e-06,
+      "loss": 3.1793,
+      "step": 6885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.3961330655960874e-06,
+      "loss": 3.8664,
+      "step": 6886
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.442930591433936e-06,
+      "loss": 1.6448,
+      "step": 6887
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.489924489779818e-06,
+      "loss": 1.5428,
+      "step": 6888
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.537114663038534e-06,
+      "loss": 3.1032,
+      "step": 6889
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.584501013207344e-06,
+      "loss": 0.9056,
+      "step": 6890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.63208344187608e-06,
+      "loss": 2.762,
+      "step": 6891
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.679861850227386e-06,
+      "loss": 2.6295,
+      "step": 6892
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.727836139036868e-06,
+      "loss": 2.5739,
+      "step": 6893
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.776006208673368e-06,
+      "loss": 2.0921,
+      "step": 6894
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.824371959099128e-06,
+      "loss": 2.4089,
+      "step": 6895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.872933289870009e-06,
+      "loss": 3.0678,
+      "step": 6896
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.921690100135679e-06,
+      "loss": 2.7111,
+      "step": 6897
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.970642288639939e-06,
+      "loss": 6.3755,
+      "step": 6898
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.0197897537206285e-06,
+      "loss": 2.9373,
+      "step": 6899
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.0691323933102175e-06,
+      "loss": 4.1847,
+      "step": 6900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.118670104935753e-06,
+      "loss": 1.8739,
+      "step": 6901
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.1684027857193185e-06,
+      "loss": 1.628,
+      "step": 6902
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.218330332377931e-06,
+      "loss": 1.1449,
+      "step": 6903
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.2684526412239545e-06,
+      "loss": 3.5379,
+      "step": 6904
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.31876960816532e-06,
+      "loss": 4.0331,
+      "step": 6905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.369281128705651e-06,
+      "loss": 3.7024,
+      "step": 6906
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.419987097944557e-06,
+      "loss": 1.8045,
+      "step": 6907
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.470887410577831e-06,
+      "loss": 2.4557,
+      "step": 6908
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.521981960897638e-06,
+      "loss": 5.0005,
+      "step": 6909
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.573270642792772e-06,
+      "loss": 0.9605,
+      "step": 6910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.624753349748847e-06,
+      "loss": 2.8736,
+      "step": 6911
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.676429974848563e-06,
+      "loss": 3.3594,
+      "step": 6912
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.728300410771871e-06,
+      "loss": 3.5062,
+      "step": 6913
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.780364549796237e-06,
+      "loss": 3.6573,
+      "step": 6914
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.83262228379683e-06,
+      "loss": 2.8748,
+      "step": 6915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.8850735042468084e-06,
+      "loss": 1.1544,
+      "step": 6916
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.937718102217472e-06,
+      "loss": 3.682,
+      "step": 6917
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.990555968378531e-06,
+      "loss": 0.4484,
+      "step": 6918
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.043586992998252e-06,
+      "loss": 1.2824,
+      "step": 6919
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.096811065943976e-06,
+      "loss": 3.8081,
+      "step": 6920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.1502280766819085e-06,
+      "loss": 1.6136,
+      "step": 6921
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.203837914277633e-06,
+      "loss": 4.872,
+      "step": 6922
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.257640467396309e-06,
+      "loss": 4.2116,
+      "step": 6923
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.311635624302837e-06,
+      "loss": 3.228,
+      "step": 6924
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.365823272862138e-06,
+      "loss": 3.9935,
+      "step": 6925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.420203300539386e-06,
+      "loss": 2.4578,
+      "step": 6926
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.474775594400207e-06,
+      "loss": 2.5662,
+      "step": 6927
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.529540041110939e-06,
+      "loss": 2.4422,
+      "step": 6928
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.584496526938889e-06,
+      "loss": 3.0195,
+      "step": 6929
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.6396449377526e-06,
+      "loss": 0.9871,
+      "step": 6930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.694985159021795e-06,
+      "loss": 3.2222,
+      "step": 6931
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.750517075818108e-06,
+      "loss": 4.6135,
+      "step": 6932
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.806240572814915e-06,
+      "loss": 2.8609,
+      "step": 6933
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.862155534287908e-06,
+      "loss": 2.8409,
+      "step": 6934
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.918261844114961e-06,
+      "loss": 2.2692,
+      "step": 6935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.974559385776637e-06,
+      "loss": 4.2299,
+      "step": 6936
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.03104804235637e-06,
+      "loss": 2.6485,
+      "step": 6937
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.087727696540659e-06,
+      "loss": 3.5643,
+      "step": 6938
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.144598230619371e-06,
+      "loss": 2.9358,
+      "step": 6939
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.20165952648595e-06,
+      "loss": 2.0942,
+      "step": 6940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.258911465637675e-06,
+      "loss": 1.6102,
+      "step": 6941
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.316353929175924e-06,
+      "loss": 3.4313,
+      "step": 6942
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.373986797806365e-06,
+      "loss": 0.1609,
+      "step": 6943
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.431809951839265e-06,
+      "loss": 2.3321,
+      "step": 6944
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.489823271189711e-06,
+      "loss": 1.6761,
+      "step": 6945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.548026635377848e-06,
+      "loss": 2.9185,
+      "step": 6946
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.606419923529174e-06,
+      "loss": 3.0609,
+      "step": 6947
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.665003014374717e-06,
+      "loss": 2.9641,
+      "step": 6948
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.723775786251375e-06,
+      "loss": 3.3282,
+      "step": 6949
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.782738117102007e-06,
+      "loss": 4.8224,
+      "step": 6950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.841889884476061e-06,
+      "loss": 2.4487,
+      "step": 6951
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.901230965529329e-06,
+      "loss": 2.5325,
+      "step": 6952
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.960761237024507e-06,
+      "loss": 3.2888,
+      "step": 6953
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.020480575331425e-06,
+      "loss": 2.7443,
+      "step": 6954
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.080388856427235e-06,
+      "loss": 2.9489,
+      "step": 6955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.140485955896705e-06,
+      "loss": 4.8878,
+      "step": 6956
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.200771748932469e-06,
+      "loss": 5.6651,
+      "step": 6957
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.261246110335309e-06,
+      "loss": 3.5091,
+      "step": 6958
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.321908914514388e-06,
+      "loss": 1.7556,
+      "step": 6959
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.382760035487492e-06,
+      "loss": 3.9948,
+      "step": 6960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.443799346881354e-06,
+      "loss": 3.5658,
+      "step": 6961
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.505026721931931e-06,
+      "loss": 2.991,
+      "step": 6962
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.566442033484412e-06,
+      "loss": 1.8757,
+      "step": 6963
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.628045153993815e-06,
+      "loss": 2.6179,
+      "step": 6964
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.689835955525283e-06,
+      "loss": 4.6756,
+      "step": 6965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.75181430975397e-06,
+      "loss": 2.502,
+      "step": 6966
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.813980087965601e-06,
+      "loss": 3.5649,
+      "step": 6967
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.87633316105665e-06,
+      "loss": 2.7117,
+      "step": 6968
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.938873399534665e-06,
+      "loss": 4.438,
+      "step": 6969
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.000160067351844e-05,
+      "loss": 6.8817,
+      "step": 6970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0064514852738393e-05,
+      "loss": 3.0311,
+      "step": 6971
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.012761580653674e-05,
+      "loss": 1.7868,
+      "step": 6972
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0190903403867846e-05,
+      "loss": 0.6913,
+      "step": 6973
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.025437751329844e-05,
+      "loss": 3.6194,
+      "step": 6974
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0318038003007923e-05,
+      "loss": 2.2786,
+      "step": 6975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0381884740788628e-05,
+      "loss": 1.2398,
+      "step": 6976
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0445917594046072e-05,
+      "loss": 2.2766,
+      "step": 6977
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.051013642979929e-05,
+      "loss": 2.4777,
+      "step": 6978
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0574541114681047e-05,
+      "loss": 2.898,
+      "step": 6979
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.063913151493816e-05,
+      "loss": 4.6986,
+      "step": 6980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0703907496431753e-05,
+      "loss": 3.8506,
+      "step": 6981
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0768868924637466e-05,
+      "loss": 4.2069,
+      "step": 6982
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0834015664646046e-05,
+      "loss": 3.0472,
+      "step": 6983
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0899347581163155e-05,
+      "loss": 3.527,
+      "step": 6984
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.0964864538509967e-05,
+      "loss": 3.3442,
+      "step": 6985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1030566400623354e-05,
+      "loss": 2.8189,
+      "step": 6986
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.109645303105622e-05,
+      "loss": 4.5545,
+      "step": 6987
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1162524292977716e-05,
+      "loss": 2.423,
+      "step": 6988
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1228780049173571e-05,
+      "loss": 1.1594,
+      "step": 6989
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.129522016204637e-05,
+      "loss": 0.2474,
+      "step": 6990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1361844493615815e-05,
+      "loss": 4.1363,
+      "step": 6991
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1428652905519054e-05,
+      "loss": 1.2634,
+      "step": 6992
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1495645259010924e-05,
+      "loss": 3.4347,
+      "step": 6993
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1562821414964353e-05,
+      "loss": 3.3431,
+      "step": 6994
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1630181233870285e-05,
+      "loss": 3.0683,
+      "step": 6995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1697724575838443e-05,
+      "loss": 2.9746,
+      "step": 6996
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1765451300597551e-05,
+      "loss": 2.7291,
+      "step": 6997
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1833361267495247e-05,
+      "loss": 2.1612,
+      "step": 6998
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1901454335498729e-05,
+      "loss": 2.2766,
+      "step": 6999
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.1969730363194887e-05,
+      "loss": 1.3814,
+      "step": 7000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2038189208790695e-05,
+      "loss": 3.2751,
+      "step": 7001
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2106830730113438e-05,
+      "loss": 3.701,
+      "step": 7002
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2175654784611024e-05,
+      "loss": 2.7929,
+      "step": 7003
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2244661229352283e-05,
+      "loss": 3.4228,
+      "step": 7004
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2313849921027277e-05,
+      "loss": 4.8254,
+      "step": 7005
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2383220715947563e-05,
+      "loss": 1.7791,
+      "step": 7006
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2452773470046541e-05,
+      "loss": 1.2266,
+      "step": 7007
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.25225080388797e-05,
+      "loss": 4.1675,
+      "step": 7008
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2592424277624947e-05,
+      "loss": 5.0717,
+      "step": 7009
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2662522041082925e-05,
+      "loss": 1.127,
+      "step": 7010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2732801183677257e-05,
+      "loss": 2.9373,
+      "step": 7011
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2803261559454926e-05,
+      "loss": 3.3733,
+      "step": 7012
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.287390302208642e-05,
+      "loss": 5.521,
+      "step": 7013
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2944725424866422e-05,
+      "loss": 2.3842,
+      "step": 7014
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3015728620713585e-05,
+      "loss": 3.6735,
+      "step": 7015
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3086912462171198e-05,
+      "loss": 1.3848,
+      "step": 7016
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3158276801407365e-05,
+      "loss": 2.3582,
+      "step": 7017
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3229821490215355e-05,
+      "loss": 2.962,
+      "step": 7018
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3301546380013897e-05,
+      "loss": 3.2351,
+      "step": 7019
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3373451321847474e-05,
+      "loss": 3.5595,
+      "step": 7020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3445536166386663e-05,
+      "loss": 1.5577,
+      "step": 7021
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3517800763928412e-05,
+      "loss": 2.7255,
+      "step": 7022
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.359024496439636e-05,
+      "loss": 2.7214,
+      "step": 7023
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3662868617341163e-05,
+      "loss": 3.9086,
+      "step": 7024
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3735671571940788e-05,
+      "loss": 2.1367,
+      "step": 7025
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3808653677000938e-05,
+      "loss": 3.583,
+      "step": 7026
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.3881814780954894e-05,
+      "loss": 2.6297,
+      "step": 7027
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.395515473186475e-05,
+      "loss": 3.3665,
+      "step": 7028
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4028673377420798e-05,
+      "loss": 3.5542,
+      "step": 7029
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4102370564942324e-05,
+      "loss": 3.4096,
+      "step": 7030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176246141377824e-05,
+      "loss": 1.6893,
+      "step": 7031
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4250299953305347e-05,
+      "loss": 2.0776,
+      "step": 7032
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4324531846932743e-05,
+      "loss": 2.1524,
+      "step": 7033
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4398941668098055e-05,
+      "loss": 3.9988,
+      "step": 7034
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4473529262269814e-05,
+      "loss": 3.8663,
+      "step": 7035
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4548294474547352e-05,
+      "loss": 1.4892,
+      "step": 7036
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.462323714966114e-05,
+      "loss": 2.4709,
+      "step": 7037
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4698357131973095e-05,
+      "loss": 1.921,
+      "step": 7038
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.47736542654769e-05,
+      "loss": 2.6013,
+      "step": 7039
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4849128393798351e-05,
+      "loss": 1.3218,
+      "step": 7040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4924779360195662e-05,
+      "loss": 4.1288,
+      "step": 7041
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5000607007559797e-05,
+      "loss": 4.7721,
+      "step": 7042
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.507661117841479e-05,
+      "loss": 2.5467,
+      "step": 7043
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5152791714918102e-05,
+      "loss": 2.446,
+      "step": 7044
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5229148458860798e-05,
+      "loss": 2.8284,
+      "step": 7045
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.530568125166829e-05,
+      "loss": 3.8283,
+      "step": 7046
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.538238993440012e-05,
+      "loss": 1.6472,
+      "step": 7047
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.545927434775063e-05,
+      "loss": 5.7088,
+      "step": 7048
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5536334332049196e-05,
+      "loss": 5.8837,
+      "step": 7049
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5613569727260592e-05,
+      "loss": 2.7132,
+      "step": 7050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5690980372985308e-05,
+      "loss": 2.7626,
+      "step": 7051
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.576856610845986e-05,
+      "loss": 1.8323,
+      "step": 7052
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5846326772557174e-05,
+      "loss": 4.8212,
+      "step": 7053
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.5924262203786866e-05,
+      "loss": 1.7216,
+      "step": 7054
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6002372240295605e-05,
+      "loss": 2.9686,
+      "step": 7055
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.608065671986746e-05,
+      "loss": 3.5943,
+      "step": 7056
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6159115479924213e-05,
+      "loss": 4.481,
+      "step": 7057
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6237748357525807e-05,
+      "loss": 3.2013,
+      "step": 7058
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6316555189370196e-05,
+      "loss": 2.8636,
+      "step": 7059
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.639553581179466e-05,
+      "loss": 3.76,
+      "step": 7060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6474690060775155e-05,
+      "loss": 3.7503,
+      "step": 7061
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6554017771927167e-05,
+      "loss": 0.5322,
+      "step": 7062
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.663351878050593e-05,
+      "loss": 3.4168,
+      "step": 7063
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6713192921406807e-05,
+      "loss": 3.1136,
+      "step": 7064
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6793040029165563e-05,
+      "loss": 3.0332,
+      "step": 7065
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.687305993795879e-05,
+      "loss": 1.9846,
+      "step": 7066
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.69532524816042e-05,
+      "loss": 1.247,
+      "step": 7067
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.703361749356097e-05,
+      "loss": 3.3889,
+      "step": 7068
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.711415480693014e-05,
+      "loss": 4.7017,
+      "step": 7069
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.719486425445487e-05,
+      "loss": 4.0992,
+      "step": 7070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7275745668520883e-05,
+      "loss": 4.9071,
+      "step": 7071
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.735679888115672e-05,
+      "loss": 1.8946,
+      "step": 7072
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7438023724034203e-05,
+      "loss": 4.4774,
+      "step": 7073
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.751942002846867e-05,
+      "loss": 2.1495,
+      "step": 7074
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7600987625419375e-05,
+      "loss": 2.3109,
+      "step": 7075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7682726345489764e-05,
+      "loss": 3.2433,
+      "step": 7076
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7764636018928172e-05,
+      "loss": 3.8264,
+      "step": 7077
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7846716475627635e-05,
+      "loss": 2.2116,
+      "step": 7078
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7928967545126597e-05,
+      "loss": 1.3895,
+      "step": 7079
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.801138905660916e-05,
+      "loss": 3.0866,
+      "step": 7080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8093980838905512e-05,
+      "loss": 1.4741,
+      "step": 7081
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8176742720492157e-05,
+      "loss": 3.2962,
+      "step": 7082
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.825967452949242e-05,
+      "loss": 1.7051,
+      "step": 7083
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.834277609367666e-05,
+      "loss": 5.2558,
+      "step": 7084
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.842604724046275e-05,
+      "loss": 1.0569,
+      "step": 7085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8509487796916336e-05,
+      "loss": 3.4807,
+      "step": 7086
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8593097589751262e-05,
+      "loss": 6.1115,
+      "step": 7087
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8676876445329915e-05,
+      "loss": 3.3221,
+      "step": 7088
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8760824189663563e-05,
+      "loss": 2.5706,
+      "step": 7089
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8844940648412736e-05,
+      "loss": 4.4884,
+      "step": 7090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.8929225646887593e-05,
+      "loss": 3.6475,
+      "step": 7091
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9013679010048257e-05,
+      "loss": 2.1876,
+      "step": 7092
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9098300562505232e-05,
+      "loss": 3.5984,
+      "step": 7093
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9183090128519708e-05,
+      "loss": 3.0531,
+      "step": 7094
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.926804753200394e-05,
+      "loss": 1.4473,
+      "step": 7095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9353172596521652e-05,
+      "loss": 1.5394,
+      "step": 7096
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9438465145288354e-05,
+      "loss": 2.9242,
+      "step": 7097
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.952392500117175e-05,
+      "loss": 1.4645,
+      "step": 7098
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9609551986692064e-05,
+      "loss": 2.4756,
+      "step": 7099
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.969534592402247e-05,
+      "loss": 3.3879,
+      "step": 7100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9781306634989372e-05,
+      "loss": 1.8024,
+      "step": 7101
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9867433941072855e-05,
+      "loss": 3.3731,
+      "step": 7102
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9953727663407028e-05,
+      "loss": 2.6862,
+      "step": 7103
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0040187622780395e-05,
+      "loss": 1.9365,
+      "step": 7104
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.012681363963621e-05,
+      "loss": 1.9634,
+      "step": 7105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0213605534072853e-05,
+      "loss": 3.827,
+      "step": 7106
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0300563125844263e-05,
+      "loss": 3.1421,
+      "step": 7107
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0387686234360115e-05,
+      "loss": 4.9305,
+      "step": 7108
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0474974678686676e-05,
+      "loss": 3.906,
+      "step": 7109
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0562428277546552e-05,
+      "loss": 3.3485,
+      "step": 7110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0650046849319438e-05,
+      "loss": 4.1818,
+      "step": 7111
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0737830212042432e-05,
+      "loss": 1.2629,
+      "step": 7112
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0825778183410395e-05,
+      "loss": 3.6725,
+      "step": 7113
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.0913890580776362e-05,
+      "loss": 3.007,
+      "step": 7114
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.100216722115186e-05,
+      "loss": 1.911,
+      "step": 7115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1090607921207338e-05,
+      "loss": 2.0634,
+      "step": 7116
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1179212497272526e-05,
+      "loss": 2.9268,
+      "step": 7117
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1267980765336835e-05,
+      "loss": 2.7459,
+      "step": 7118
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1356912541049713e-05,
+      "loss": 1.4544,
+      "step": 7119
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1446007639721054e-05,
+      "loss": 4.6406,
+      "step": 7120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1535265876321552e-05,
+      "loss": 0.8974,
+      "step": 7121
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.162468706548311e-05,
+      "loss": 0.7482,
+      "step": 7122
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.171427102149923e-05,
+      "loss": 2.7804,
+      "step": 7123
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1804017558325385e-05,
+      "loss": 1.4088,
+      "step": 7124
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1893926489579373e-05,
+      "loss": 2.4934,
+      "step": 7125
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.1983997628541774e-05,
+      "loss": 2.2613,
+      "step": 7126
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.2074230788156292e-05,
+      "loss": 1.9764,
+      "step": 7127
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.2164625781030147e-05,
+      "loss": 4.8304,
+      "step": 7128
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.225518241943446e-05,
+      "loss": 3.2233,
+      "step": 7129
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.2345900515304686e-05,
+      "loss": 2.4035,
+      "step": 7130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.2436779880240943e-05,
+      "loss": 2.7467,
+      "step": 7131
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.2527820325508442e-05,
+      "loss": 1.3478,
+      "step": 7132
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.2619021662037854e-05,
+      "loss": 4.2224,
+      "step": 7133
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.271038370042573e-05,
+      "loss": 2.2916,
+      "step": 7134
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.2801906250934902e-05,
+      "loss": 4.5555,
+      "step": 7135
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.289358912349481e-05,
+      "loss": 2.9917,
+      "step": 7136
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.2985432127701976e-05,
+      "loss": 2.0343,
+      "step": 7137
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3077435072820364e-05,
+      "loss": 2.7605,
+      "step": 7138
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.316959776778165e-05,
+      "loss": 2.8276,
+      "step": 7139
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3261920021186113e-05,
+      "loss": 4.122,
+      "step": 7140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3354401641302303e-05,
+      "loss": 1.7843,
+      "step": 7141
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3447042436067922e-05,
+      "loss": 1.504,
+      "step": 7142
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3539842213090114e-05,
+      "loss": 4.5712,
+      "step": 7143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3632800779645847e-05,
+      "loss": 2.7974,
+      "step": 7144
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.372591794268233e-05,
+      "loss": 3.6164,
+      "step": 7145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3819193508817382e-05,
+      "loss": 1.528,
+      "step": 7146
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.3912627284339893e-05,
+      "loss": 3.5292,
+      "step": 7147
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.4006219075210145e-05,
+      "loss": 2.3547,
+      "step": 7148
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.4099968687060305e-05,
+      "loss": 3.2134,
+      "step": 7149
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.4193875925194743e-05,
+      "loss": 5.1041,
+      "step": 7150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.428794059459051e-05,
+      "loss": 2.6896,
+      "step": 7151
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.4382162499897688e-05,
+      "loss": 2.1872,
+      "step": 7152
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.4476541445439828e-05,
+      "loss": 1.8063,
+      "step": 7153
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.4571077235214334e-05,
+      "loss": 1.9441,
+      "step": 7154
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.4665769672892913e-05,
+      "loss": 3.8448,
+      "step": 7155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.4760618561821915e-05,
+      "loss": 3.7173,
+      "step": 7156
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.485562370502278e-05,
+      "loss": 3.9239,
+      "step": 7157
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.495078490519248e-05,
+      "loss": 1.7064,
+      "step": 7158
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.5046101964703862e-05,
+      "loss": 3.0917,
+      "step": 7159
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.5141574685606095e-05,
+      "loss": 2.3562,
+      "step": 7160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.52372028696251e-05,
+      "loss": 2.7305,
+      "step": 7161
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.5332986318163886e-05,
+      "loss": 2.9354,
+      "step": 7162
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.542892483230307e-05,
+      "loss": 4.3982,
+      "step": 7163
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.5525018212801222e-05,
+      "loss": 0.3259,
+      "step": 7164
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.5621266260095255e-05,
+      "loss": 2.5033,
+      "step": 7165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.5717668774300917e-05,
+      "loss": 3.2989,
+      "step": 7166
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.5814225555213156e-05,
+      "loss": 3.2094,
+      "step": 7167
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.591093640230653e-05,
+      "loss": 1.4348,
+      "step": 7168
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.6007801114735652e-05,
+      "loss": 2.5092,
+      "step": 7169
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.6104819491335596e-05,
+      "loss": 1.6903,
+      "step": 7170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.6201991330622167e-05,
+      "loss": 2.5029,
+      "step": 7171
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.6299316430792864e-05,
+      "loss": 3.6007,
+      "step": 7172
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.6396794589726504e-05,
+      "loss": 1.4336,
+      "step": 7173
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.6494425604984175e-05,
+      "loss": 2.6639,
+      "step": 7174
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.659220927380952e-05,
+      "loss": 1.8069,
+      "step": 7175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.669014539312915e-05,
+      "loss": 2.5232,
+      "step": 7176
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.678823375955307e-05,
+      "loss": 3.9744,
+      "step": 7177
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.688647416937512e-05,
+      "loss": 3.6056,
+      "step": 7178
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.6984866418573362e-05,
+      "loss": 2.2333,
+      "step": 7179
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.708341030281054e-05,
+      "loss": 2.3334,
+      "step": 7180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.7182105617434484e-05,
+      "loss": 2.0129,
+      "step": 7181
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.7280952157478523e-05,
+      "loss": 2.2343,
+      "step": 7182
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.7379949717661955e-05,
+      "loss": 3.7261,
+      "step": 7183
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.747909809239042e-05,
+      "loss": 1.868,
+      "step": 7184
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.757839707575637e-05,
+      "loss": 2.8522,
+      "step": 7185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.7677846461539458e-05,
+      "loss": 4.3255,
+      "step": 7186
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.7777446043207024e-05,
+      "loss": 3.3539,
+      "step": 7187
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.787719561391444e-05,
+      "loss": 3.1447,
+      "step": 7188
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.7977094966505624e-05,
+      "loss": 4.2011,
+      "step": 7189
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8077143893513413e-05,
+      "loss": 3.0111,
+      "step": 7190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8177342187160017e-05,
+      "loss": 2.506,
+      "step": 7191
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8277689639357452e-05,
+      "loss": 3.2347,
+      "step": 7192
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8378186041707967e-05,
+      "loss": 1.5999,
+      "step": 7193
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.847883118550447e-05,
+      "loss": 3.4985,
+      "step": 7194
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.857962486173098e-05,
+      "loss": 3.055,
+      "step": 7195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.868056686106303e-05,
+      "loss": 1.2695,
+      "step": 7196
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8781656973868144e-05,
+      "loss": 4.7218,
+      "step": 7197
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8882894990206233e-05,
+      "loss": 3.6258,
+      "step": 7198
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8984280699830068e-05,
+      "loss": 1.66,
+      "step": 7199
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.9085813892185652e-05,
+      "loss": 4.3492,
+      "step": 7200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.918749435641276e-05,
+      "loss": 2.0515,
+      "step": 7201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.9289321881345288e-05,
+      "loss": 2.9134,
+      "step": 7202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.939129625551158e-05,
+      "loss": 1.5295,
+      "step": 7203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.9493417267135416e-05,
+      "loss": 3.168,
+      "step": 7204
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.959568470413565e-05,
+      "loss": 0.9581,
+      "step": 7205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.969809835412719e-05,
+      "loss": 3.1762,
+      "step": 7206
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.9800658004421267e-05,
+      "loss": 4.5178,
+      "step": 7207
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.9903363442025934e-05,
+      "loss": 2.7205,
+      "step": 7208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.000621445364644e-05,
+      "loss": 2.6128,
+      "step": 7209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0109210825685762e-05,
+      "loss": 0.785,
+      "step": 7210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0212352344244955e-05,
+      "loss": 4.2069,
+      "step": 7211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0315638795123657e-05,
+      "loss": 3.5596,
+      "step": 7212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.0419069963820522e-05,
+      "loss": 3.9338,
+      "step": 7213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.052264563553364e-05,
+      "loss": 1.8955,
+      "step": 7214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.062636559516102e-05,
+      "loss": 1.8316,
+      "step": 7215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.073022962730102e-05,
+      "loss": 2.0837,
+      "step": 7216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.083423751625278e-05,
+      "loss": 3.4304,
+      "step": 7217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.093838904601669e-05,
+      "loss": 1.1674,
+      "step": 7218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.104268400029484e-05,
+      "loss": 2.814,
+      "step": 7219
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.114712216249145e-05,
+      "loss": 3.8445,
+      "step": 7220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.125170331571333e-05,
+      "loss": 2.1704,
+      "step": 7221
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.1356427242770346e-05,
+      "loss": 4.7905,
+      "step": 7222
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.146129372617586e-05,
+      "loss": 4.9074,
+      "step": 7223
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.156630254814714e-05,
+      "loss": 2.3863,
+      "step": 7224
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.167145349060591e-05,
+      "loss": 1.5215,
+      "step": 7225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.177674633517868e-05,
+      "loss": 1.9703,
+      "step": 7226
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.188218086319731e-05,
+      "loss": 2.908,
+      "step": 7227
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.19877568556994e-05,
+      "loss": 4.0079,
+      "step": 7228
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.2093474093428767e-05,
+      "loss": 2.0809,
+      "step": 7229
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.219933235683589e-05,
+      "loss": 1.7669,
+      "step": 7230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.2305331426078364e-05,
+      "loss": 3.9998,
+      "step": 7231
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.2411471081021385e-05,
+      "loss": 3.2261,
+      "step": 7232
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.251775110123817e-05,
+      "loss": 1.8504,
+      "step": 7233
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.26241712660103e-05,
+      "loss": 0.4785,
+      "step": 7234
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.2730731354328715e-05,
+      "loss": 3.8064,
+      "step": 7235
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.283743114489336e-05,
+      "loss": 2.7758,
+      "step": 7236
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.2944270416114185e-05,
+      "loss": 1.362,
+      "step": 7237
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.305124894611148e-05,
+      "loss": 3.0543,
+      "step": 7238
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.315836651271632e-05,
+      "loss": 1.0644,
+      "step": 7239
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.3265622893471036e-05,
+      "loss": 2.1788,
+      "step": 7240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.337301786562967e-05,
+      "loss": 2.6521,
+      "step": 7241
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.348055120615846e-05,
+      "loss": 2.9181,
+      "step": 7242
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.358822269173627e-05,
+      "loss": 4.0795,
+      "step": 7243
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.369603209875507e-05,
+      "loss": 2.033,
+      "step": 7244
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.38039792033204e-05,
+      "loss": 3.2185,
+      "step": 7245
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.3912063781251856e-05,
+      "loss": 3.9921,
+      "step": 7246
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4020285608083505e-05,
+      "loss": 1.7884,
+      "step": 7247
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4128644459064384e-05,
+      "loss": 2.9007,
+      "step": 7248
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.423714010915898e-05,
+      "loss": 4.9704,
+      "step": 7249
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4345772333047686e-05,
+      "loss": 3.2912,
+      "step": 7250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4454540905127233e-05,
+      "loss": 2.5007,
+      "step": 7251
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.456344559951122e-05,
+      "loss": 1.1363,
+      "step": 7252
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.467248619003054e-05,
+      "loss": 4.0139,
+      "step": 7253
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.478166245023384e-05,
+      "loss": 1.5051,
+      "step": 7254
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.489097415338807e-05,
+      "loss": 2.9901,
+      "step": 7255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.500042107247884e-05,
+      "loss": 2.6882,
+      "step": 7256
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.5110002980210975e-05,
+      "loss": 1.3475,
+      "step": 7257
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.521971964900895e-05,
+      "loss": 2.6192,
+      "step": 7258
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.53295708510174e-05,
+      "loss": 3.6434,
+      "step": 7259
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.5439556358101534e-05,
+      "loss": 1.1889,
+      "step": 7260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.554967594184765e-05,
+      "loss": 0.4598,
+      "step": 7261
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.5659929373563613e-05,
+      "loss": 2.3517,
+      "step": 7262
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.5770316424279314e-05,
+      "loss": 3.5061,
+      "step": 7263
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.5880836864747126e-05,
+      "loss": 2.6685,
+      "step": 7264
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.5991490465442445e-05,
+      "loss": 1.6587,
+      "step": 7265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.610227699656395e-05,
+      "loss": 3.4763,
+      "step": 7266
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.621319622803468e-05,
+      "loss": 1.6349,
+      "step": 7267
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.63242479295017e-05,
+      "loss": 4.8813,
+      "step": 7268
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.6435431870337045e-05,
+      "loss": 1.4809,
+      "step": 7269
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.654674781963818e-05,
+      "loss": 4.207,
+      "step": 7270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.66581955462284e-05,
+      "loss": 1.4581,
+      "step": 7271
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.676977481865731e-05,
+      "loss": 3.4044,
+      "step": 7272
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.6881485405201346e-05,
+      "loss": 4.0464,
+      "step": 7273
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.6993327073864225e-05,
+      "loss": 2.2045,
+      "step": 7274
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7105299592377454e-05,
+      "loss": 2.9085,
+      "step": 7275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.721740272820079e-05,
+      "loss": 3.7036,
+      "step": 7276
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7329636248522714e-05,
+      "loss": 3.019,
+      "step": 7277
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.744199992026094e-05,
+      "loss": 1.5796,
+      "step": 7278
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.755449351006291e-05,
+      "loss": 3.4199,
+      "step": 7279
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.76671167843062e-05,
+      "loss": 1.9336,
+      "step": 7280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.777986950909912e-05,
+      "loss": 3.1146,
+      "step": 7281
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7892751450281126e-05,
+      "loss": 2.387,
+      "step": 7282
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.80057623734233e-05,
+      "loss": 5.8558,
+      "step": 7283
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.81189020438289e-05,
+      "loss": 3.7147,
+      "step": 7284
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.8232170226533745e-05,
+      "loss": 1.2344,
+      "step": 7285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.834556668630684e-05,
+      "loss": 2.3359,
+      "step": 7286
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.845909118765073e-05,
+      "loss": 2.0745,
+      "step": 7287
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.857274349480208e-05,
+      "loss": 1.9431,
+      "step": 7288
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.86865233717321e-05,
+      "loss": 3.1448,
+      "step": 7289
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.880043058214712e-05,
+      "loss": 4.9841,
+      "step": 7290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.891446488948898e-05,
+      "loss": 2.3823,
+      "step": 7291
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.902862605693559e-05,
+      "loss": 4.1912,
+      "step": 7292
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.9142913847401417e-05,
+      "loss": 0.261,
+      "step": 7293
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.925732802353792e-05,
+      "loss": 2.3959,
+      "step": 7294
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.937186834773413e-05,
+      "loss": 3.2725,
+      "step": 7295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.948653458211707e-05,
+      "loss": 3.0085,
+      "step": 7296
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.9601326488552145e-05,
+      "loss": 2.5858,
+      "step": 7297
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.971624382864421e-05,
+      "loss": 3.192,
+      "step": 7298
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.983128636373715e-05,
+      "loss": 1.9686,
+      "step": 7299
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.9946453854915056e-05,
+      "loss": 2.689,
+      "step": 7300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.006174606300248e-05,
+      "loss": 2.8211,
+      "step": 7301
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.0177162748564955e-05,
+      "loss": 0.848,
+      "step": 7302
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.0292703671909525e-05,
+      "loss": 1.0567,
+      "step": 7303
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.040836859308522e-05,
+      "loss": 1.0206,
+      "step": 7304
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.052415727188356e-05,
+      "loss": 4.1188,
+      "step": 7305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.0640069467839024e-05,
+      "loss": 2.3074,
+      "step": 7306
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.07561049402296e-05,
+      "loss": 2.415,
+      "step": 7307
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.087226344807726e-05,
+      "loss": 2.77,
+      "step": 7308
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.098854475014845e-05,
+      "loss": 2.3785,
+      "step": 7309
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.110494860495461e-05,
+      "loss": 3.02,
+      "step": 7310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.122147477075266e-05,
+      "loss": 3.701,
+      "step": 7311
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.1338123005545504e-05,
+      "loss": 3.1993,
+      "step": 7312
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.145489306708256e-05,
+      "loss": 3.7014,
+      "step": 7313
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.1571784712860215e-05,
+      "loss": 0.8578,
+      "step": 7314
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.168879770012237e-05,
+      "loss": 2.9942,
+      "step": 7315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.180593178586093e-05,
+      "loss": 2.5619,
+      "step": 7316
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.19231867268163e-05,
+      "loss": 1.9173,
+      "step": 7317
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.20405622794779e-05,
+      "loss": 3.3234,
+      "step": 7318
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.215805820008467e-05,
+      "loss": 3.0037,
+      "step": 7319
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.227567424462556e-05,
+      "loss": 3.5631,
+      "step": 7320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.239341016884009e-05,
+      "loss": 3.1438,
+      "step": 7321
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.251126572821876e-05,
+      "loss": 3.3092,
+      "step": 7322
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.2629240678003665e-05,
+      "loss": 4.0896,
+      "step": 7323
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.274733477318895e-05,
+      "loss": 1.2948,
+      "step": 7324
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.286554776852129e-05,
+      "loss": 1.6925,
+      "step": 7325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.298387941850046e-05,
+      "loss": 0.931,
+      "step": 7326
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.310232947737982e-05,
+      "loss": 3.9676,
+      "step": 7327
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.322089769916681e-05,
+      "loss": 4.0103,
+      "step": 7328
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.333958383762332e-05,
+      "loss": 2.1524,
+      "step": 7329
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.345838764626683e-05,
+      "loss": 1.0584,
+      "step": 7330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.357730887836997e-05,
+      "loss": 1.5778,
+      "step": 7331
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.369634728696167e-05,
+      "loss": 3.1257,
+      "step": 7332
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.381550262482751e-05,
+      "loss": 2.4344,
+      "step": 7333
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.393477464451024e-05,
+      "loss": 3.2737,
+      "step": 7334
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.40541630983103e-05,
+      "loss": 3.4204,
+      "step": 7335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.4173667738286294e-05,
+      "loss": 2.008,
+      "step": 7336
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.429328831625558e-05,
+      "loss": 0.5129,
+      "step": 7337
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.441302458379467e-05,
+      "loss": 3.535,
+      "step": 7338
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.45328762922399e-05,
+      "loss": 5.0814,
+      "step": 7339
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.4652843192687796e-05,
+      "loss": 2.5204,
+      "step": 7340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.47729250359957e-05,
+      "loss": 3.0454,
+      "step": 7341
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.489312157278223e-05,
+      "loss": 3.9274,
+      "step": 7342
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.501343255342779e-05,
+      "loss": 2.2013,
+      "step": 7343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.5133857728075166e-05,
+      "loss": 4.1014,
+      "step": 7344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.525439684662993e-05,
+      "loss": 3.3635,
+      "step": 7345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.5375049658760914e-05,
+      "loss": 2.3669,
+      "step": 7346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.549581591390141e-05,
+      "loss": 4.654,
+      "step": 7347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.561669536124823e-05,
+      "loss": 4.3131,
+      "step": 7348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.57376877497637e-05,
+      "loss": 2.499,
+      "step": 7349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.5858792828175467e-05,
+      "loss": 2.9962,
+      "step": 7350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.598001034497712e-05,
+      "loss": 2.6509,
+      "step": 7351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.610134004842875e-05,
+      "loss": 1.26,
+      "step": 7352
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.622278168655747e-05,
+      "loss": 2.5862,
+      "step": 7353
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.6344335007157935e-05,
+      "loss": 2.8103,
+      "step": 7354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.6465999757792845e-05,
+      "loss": 6.1184,
+      "step": 7355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.6587775685793497e-05,
+      "loss": 1.9121,
+      "step": 7356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.67096625382603e-05,
+      "loss": 1.8639,
+      "step": 7357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.68316600620633e-05,
+      "loss": 2.2884,
+      "step": 7358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.6953768003842694e-05,
+      "loss": 5.1608,
+      "step": 7359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.707598611000923e-05,
+      "loss": 2.5284,
+      "step": 7360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.719831412674531e-05,
+      "loss": 3.575,
+      "step": 7361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.732075180000465e-05,
+      "loss": 4.5375,
+      "step": 7362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.7443298875513354e-05,
+      "loss": 5.2626,
+      "step": 7363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.7565955098770345e-05,
+      "loss": 5.3782,
+      "step": 7364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.768872021504785e-05,
+      "loss": 4.1351,
+      "step": 7365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.781159396939201e-05,
+      "loss": 4.6134,
+      "step": 7366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.793457610662326e-05,
+      "loss": 3.3761,
+      "step": 7367
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.805766637133703e-05,
+      "loss": 2.5201,
+      "step": 7368
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.818086450790414e-05,
+      "loss": 2.6834,
+      "step": 7369
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.830417026047143e-05,
+      "loss": 3.1799,
+      "step": 7370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.842758337296222e-05,
+      "loss": 1.0602,
+      "step": 7371
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.8551103589076854e-05,
+      "loss": 2.5457,
+      "step": 7372
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.8674730652293275e-05,
+      "loss": 5.7712,
+      "step": 7373
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.879846430586755e-05,
+      "loss": 4.5566,
+      "step": 7374
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.892230429283433e-05,
+      "loss": 2.6596,
+      "step": 7375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.9046250356007475e-05,
+      "loss": 3.8401,
+      "step": 7376
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.9170302237980525e-05,
+      "loss": 4.7656,
+      "step": 7377
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.9294459681127126e-05,
+      "loss": 1.5038,
+      "step": 7378
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.941872242760232e-05,
+      "loss": 5.3198,
+      "step": 7379
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.9543090219341495e-05,
+      "loss": 2.7682,
+      "step": 7380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.966756279806255e-05,
+      "loss": 3.0481,
+      "step": 7381
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.9792139905265575e-05,
+      "loss": 3.2476,
+      "step": 7382
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.9916821282233615e-05,
+      "loss": 3.3536,
+      "step": 7383
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.004160667003312e-05,
+      "loss": 1.9646,
+      "step": 7384
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.0166495809514613e-05,
+      "loss": 2.1065,
+      "step": 7385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.029148844131306e-05,
+      "loss": 3.8696,
+      "step": 7386
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.041658430584856e-05,
+      "loss": 1.4796,
+      "step": 7387
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.0541783143326795e-05,
+      "loss": 1.8757,
+      "step": 7388
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.06670846937396e-05,
+      "loss": 2.7143,
+      "step": 7389
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.0792488696865524e-05,
+      "loss": 2.7566,
+      "step": 7390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.091799489227029e-05,
+      "loss": 2.1658,
+      "step": 7391
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.104360301930732e-05,
+      "loss": 0.9106,
+      "step": 7392
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.116931281711872e-05,
+      "loss": 4.0418,
+      "step": 7393
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.1295124024635074e-05,
+      "loss": 3.5292,
+      "step": 7394
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.1421036380576436e-05,
+      "loss": 0.774,
+      "step": 7395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.154704962345287e-05,
+      "loss": 5.5233,
+      "step": 7396
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.1673163491564855e-05,
+      "loss": 2.1564,
+      "step": 7397
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.179937772300396e-05,
+      "loss": 3.1278,
+      "step": 7398
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.192569205565327e-05,
+      "loss": 4.3599,
+      "step": 7399
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.205210622718804e-05,
+      "loss": 3.2019,
+      "step": 7400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.2178619975076107e-05,
+      "loss": 2.5894,
+      "step": 7401
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.2305233036578575e-05,
+      "loss": 3.1244,
+      "step": 7402
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.2431945148750293e-05,
+      "loss": 1.8514,
+      "step": 7403
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.255875604844036e-05,
+      "loss": 1.8974,
+      "step": 7404
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.268566547229276e-05,
+      "loss": 1.7358,
+      "step": 7405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.281267315674685e-05,
+      "loss": 4.4093,
+      "step": 7406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.293977883803792e-05,
+      "loss": 3.7528,
+      "step": 7407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.306698225219776e-05,
+      "loss": 4.1499,
+      "step": 7408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.3194283135055026e-05,
+      "loss": 2.0024,
+      "step": 7409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.3321681222236396e-05,
+      "loss": 2.4909,
+      "step": 7410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.3449176249166435e-05,
+      "loss": 2.3782,
+      "step": 7411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.357676795106799e-05,
+      "loss": 4.6454,
+      "step": 7412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.370445606296366e-05,
+      "loss": 3.6014,
+      "step": 7413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.383224031967566e-05,
+      "loss": 1.791,
+      "step": 7414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.3960120455826545e-05,
+      "loss": 0.235,
+      "step": 7415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.40880962058397e-05,
+      "loss": 3.9828,
+      "step": 7416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.421616730394e-05,
+      "loss": 3.1225,
+      "step": 7417
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.4344333484154286e-05,
+      "loss": 1.5716,
+      "step": 7418
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.447259448031191e-05,
+      "loss": 3.5246,
+      "step": 7419
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.460095002604537e-05,
+      "loss": 2.0674,
+      "step": 7420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.4729399854790755e-05,
+      "loss": 5.1457,
+      "step": 7421
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.485794369978836e-05,
+      "loss": 3.0128,
+      "step": 7422
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.49865812940831e-05,
+      "loss": 2.5241,
+      "step": 7423
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.511531237052565e-05,
+      "loss": 1.8122,
+      "step": 7424
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.524413666177206e-05,
+      "loss": 2.9421,
+      "step": 7425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.537305390028498e-05,
+      "loss": 4.454,
+      "step": 7426
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.5502063818334004e-05,
+      "loss": 4.478,
+      "step": 7427
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.5631166147996284e-05,
+      "loss": 2.5722,
+      "step": 7428
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.5760360621157015e-05,
+      "loss": 1.6429,
+      "step": 7429
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.5889646969510056e-05,
+      "loss": 1.3726,
+      "step": 7430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.6019024924558464e-05,
+      "loss": 2.0279,
+      "step": 7431
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.614849421761505e-05,
+      "loss": 3.4053,
+      "step": 7432
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 5.627805457980293e-05,
+      "loss": 1.057,
+      "step": 7433
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.6407705742056094e-05,
+      "loss": 1.4442,
+      "step": 7434
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.653744743511996e-05,
+      "loss": 3.4824,
+      "step": 7435
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.6667279389551945e-05,
+      "loss": 1.8616,
+      "step": 7436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.679720133572202e-05,
+      "loss": 4.7303,
+      "step": 7437
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.6927213003813215e-05,
+      "loss": 3.7913,
+      "step": 7438
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.7057314123822304e-05,
+      "loss": 0.8391,
+      "step": 7439
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.7187504425560234e-05,
+      "loss": 2.3281,
+      "step": 7440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.731778363865261e-05,
+      "loss": 2.6761,
+      "step": 7441
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.744815149254086e-05,
+      "loss": 1.6713,
+      "step": 7442
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.7578607716482006e-05,
+      "loss": 1.8633,
+      "step": 7443
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.770915203954923e-05,
+      "loss": 4.1386,
+      "step": 7444
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.7839784190633226e-05,
+      "loss": 2.8155,
+      "step": 7445
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.797050389844214e-05,
+      "loss": 3.7182,
+      "step": 7446
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.810131089150227e-05,
+      "loss": 4.2495,
+      "step": 7447
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.823220489815866e-05,
+      "loss": 1.6206,
+      "step": 7448
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.8363185646575644e-05,
+      "loss": 4.836,
+      "step": 7449
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.8494252864737395e-05,
+      "loss": 1.2459,
+      "step": 7450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.862540628044855e-05,
+      "loss": 3.1727,
+      "step": 7451
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.875664562133467e-05,
+      "loss": 2.1634,
+      "step": 7452
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.8887970614842935e-05,
+      "loss": 3.6382,
+      "step": 7453
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.9019380988242614e-05,
+      "loss": 2.2424,
+      "step": 7454
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.9150876468625495e-05,
+      "loss": 4.8982,
+      "step": 7455
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.928245678290711e-05,
+      "loss": 1.3181,
+      "step": 7456
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.941412165782635e-05,
+      "loss": 2.2846,
+      "step": 7457
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.9545870819946616e-05,
+      "loss": 3.8857,
+      "step": 7458
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.9677703995656283e-05,
+      "loss": 1.146,
+      "step": 7459
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.980962091116927e-05,
+      "loss": 3.2383,
+      "step": 7460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.994162129252555e-05,
+      "loss": 2.9336,
+      "step": 7461
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.007370486559178e-05,
+      "loss": 4.0949,
+      "step": 7462
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.020587135606184e-05,
+      "loss": 3.8167,
+      "step": 7463
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.033812048945742e-05,
+      "loss": 0.8515,
+      "step": 7464
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0470451991128575e-05,
+      "loss": 3.3243,
+      "step": 7465
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.06028655862543e-05,
+      "loss": 4.13,
+      "step": 7466
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.07353609998431e-05,
+      "loss": 3.7515,
+      "step": 7467
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.0867937956733534e-05,
+      "loss": 3.3372,
+      "step": 7468
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.10005961815949e-05,
+      "loss": 3.067,
+      "step": 7469
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.113333539892762e-05,
+      "loss": 1.3827,
+      "step": 7470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.1266155333064e-05,
+      "loss": 3.7242,
+      "step": 7471
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.139905570816848e-05,
+      "loss": 1.8781,
+      "step": 7472
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.1532036248239e-05,
+      "loss": 3.0118,
+      "step": 7473
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.166509667710645e-05,
+      "loss": 1.9378,
+      "step": 7474
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.179823671843621e-05,
+      "loss": 4.9086,
+      "step": 7475
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.19314560957278e-05,
+      "loss": 0.9614,
+      "step": 7476
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.206475453231646e-05,
+      "loss": 2.9364,
+      "step": 7477
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.21981317513731e-05,
+      "loss": 1.0498,
+      "step": 7478
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.233158747590502e-05,
+      "loss": 2.5662,
+      "step": 7479
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.246512142875645e-05,
+      "loss": 3.551,
+      "step": 7480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.25987333326092e-05,
+      "loss": 4.554,
+      "step": 7481
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.273242290998315e-05,
+      "loss": 4.3864,
+      "step": 7482
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.286618988323692e-05,
+      "loss": 3.3926,
+      "step": 7483
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.300003397456836e-05,
+      "loss": 3.0147,
+      "step": 7484
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.313395490601519e-05,
+      "loss": 1.5492,
+      "step": 7485
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.326795239945535e-05,
+      "loss": 2.2497,
+      "step": 7486
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.340202617660831e-05,
+      "loss": 0.2829,
+      "step": 7487
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.353617595903461e-05,
+      "loss": 3.5117,
+      "step": 7488
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.36704014681371e-05,
+      "loss": 2.9926,
+      "step": 7489
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.380470242516139e-05,
+      "loss": 1.5299,
+      "step": 7490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.393907855119636e-05,
+      "loss": 1.0042,
+      "step": 7491
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.407352956717483e-05,
+      "loss": 3.1643,
+      "step": 7492
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.420805519387405e-05,
+      "loss": 5.8248,
+      "step": 7493
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.434265515191635e-05,
+      "loss": 2.5476,
+      "step": 7494
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.447732916176969e-05,
+      "loss": 0.8213,
+      "step": 7495
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.461207694374822e-05,
+      "loss": 3.82,
+      "step": 7496
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.474689821801289e-05,
+      "loss": 2.4118,
+      "step": 7497
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.488179270457205e-05,
+      "loss": 4.6692,
+      "step": 7498
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.501676012328197e-05,
+      "loss": 1.6932,
+      "step": 7499
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.515180019384748e-05,
+      "loss": 3.5357,
+      "step": 7500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.528691263582252e-05,
+      "loss": 2.8783,
+      "step": 7501
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.542209716861073e-05,
+      "loss": 3.1647,
+      "step": 7502
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.555735351146604e-05,
+      "loss": 2.9971,
+      "step": 7503
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.569268138349309e-05,
+      "loss": 1.6223,
+      "step": 7504
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.582808050364846e-05,
+      "loss": 2.3151,
+      "step": 7505
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.596355059074027e-05,
+      "loss": 3.0456,
+      "step": 7506
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.609909136342955e-05,
+      "loss": 3.5626,
+      "step": 7507
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.623470254023012e-05,
+      "loss": 2.3654,
+      "step": 7508
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.637038383951e-05,
+      "loss": 1.997,
+      "step": 7509
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.650613497949147e-05,
+      "loss": 4.9133,
+      "step": 7510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.664195567825175e-05,
+      "loss": 4.2399,
+      "step": 7511
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.677784565372359e-05,
+      "loss": 1.3935,
+      "step": 7512
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.691380462369589e-05,
+      "loss": 2.4933,
+      "step": 7513
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.704983230581429e-05,
+      "loss": 2.4092,
+      "step": 7514
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.718592841758164e-05,
+      "loss": 2.5005,
+      "step": 7515
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.732209267635878e-05,
+      "loss": 2.0066,
+      "step": 7516
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.745832479936498e-05,
+      "loss": 2.2597,
+      "step": 7517
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.759462450367837e-05,
+      "loss": 1.9842,
+      "step": 7518
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.77309915062373e-05,
+      "loss": 2.8199,
+      "step": 7519
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.786742552383982e-05,
+      "loss": 3.8905,
+      "step": 7520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.800392627314497e-05,
+      "loss": 1.2725,
+      "step": 7521
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.81404934706732e-05,
+      "loss": 1.4335,
+      "step": 7522
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.8277126832807e-05,
+      "loss": 2.7514,
+      "step": 7523
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.841382607579141e-05,
+      "loss": 3.7853,
+      "step": 7524
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.855059091573466e-05,
+      "loss": 1.0164,
+      "step": 7525
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.868742106860876e-05,
+      "loss": 4.5853,
+      "step": 7526
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.882431625025009e-05,
+      "loss": 3.5316,
+      "step": 7527
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.896127617635994e-05,
+      "loss": 3.6182,
+      "step": 7528
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.909830056250522e-05,
+      "loss": 2.8984,
+      "step": 7529
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.923538912411888e-05,
+      "loss": 2.0096,
+      "step": 7530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.937254157650065e-05,
+      "loss": 0.6347,
+      "step": 7531
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.950975763481756e-05,
+      "loss": 1.6931,
+      "step": 7532
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.964703701410452e-05,
+      "loss": 4.2704,
+      "step": 7533
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.978437942926499e-05,
+      "loss": 1.0915,
+      "step": 7534
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.99217845950715e-05,
+      "loss": 3.6624,
+      "step": 7535
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.005925222616605e-05,
+      "loss": 3.0065,
+      "step": 7536
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.019678203706147e-05,
+      "loss": 4.5218,
+      "step": 7537
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.033437374214094e-05,
+      "loss": 2.3211,
+      "step": 7538
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.047202705565943e-05,
+      "loss": 2.5245,
+      "step": 7539
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.060974169174346e-05,
+      "loss": 1.1079,
+      "step": 7540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.074751736439272e-05,
+      "loss": 3.8158,
+      "step": 7541
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.088535378747995e-05,
+      "loss": 3.4809,
+      "step": 7542
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.102325067475172e-05,
+      "loss": 0.7585,
+      "step": 7543
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.116120773982906e-05,
+      "loss": 4.4648,
+      "step": 7544
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.129922469620802e-05,
+      "loss": 0.9522,
+      "step": 7545
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.143730125726023e-05,
+      "loss": 0.5013,
+      "step": 7546
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.157543713623357e-05,
+      "loss": 2.9361,
+      "step": 7547
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.171363204625276e-05,
+      "loss": 2.0157,
+      "step": 7548
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.18518857003197e-05,
+      "loss": 2.38,
+      "step": 7549
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.199019781131482e-05,
+      "loss": 1.8278,
+      "step": 7550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.212856809199665e-05,
+      "loss": 1.9947,
+      "step": 7551
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.226699625500306e-05,
+      "loss": 0.4284,
+      "step": 7552
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.240548201285176e-05,
+      "loss": 2.6519,
+      "step": 7553
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.254402507794082e-05,
+      "loss": 3.3884,
+      "step": 7554
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.268262516254928e-05,
+      "loss": 2.2375,
+      "step": 7555
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.282128197883778e-05,
+      "loss": 2.1707,
+      "step": 7556
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.295999523884915e-05,
+      "loss": 3.202,
+      "step": 7557
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.309876465450902e-05,
+      "loss": 3.5614,
+      "step": 7558
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.323758993762633e-05,
+      "loss": 4.3441,
+      "step": 7559
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.337647079989405e-05,
+      "loss": 4.4004,
+      "step": 7560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.351540695288972e-05,
+      "loss": 1.738,
+      "step": 7561
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.365439810807604e-05,
+      "loss": 4.4046,
+      "step": 7562
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.379344397680149e-05,
+      "loss": 2.7928,
+      "step": 7563
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.393254427030094e-05,
+      "loss": 2.2996,
+      "step": 7564
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.407169869969619e-05,
+      "loss": 1.5656,
+      "step": 7565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.421090697599666e-05,
+      "loss": 3.2972,
+      "step": 7566
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.435016881009975e-05,
+      "loss": 2.4252,
+      "step": 7567
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.448948391279213e-05,
+      "loss": 3.2892,
+      "step": 7568
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.462885199474939e-05,
+      "loss": 0.6808,
+      "step": 7569
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.476827276653717e-05,
+      "loss": 2.1502,
+      "step": 7570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.490774593861194e-05,
+      "loss": 5.2307,
+      "step": 7571
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.50472712213208e-05,
+      "loss": 3.8455,
+      "step": 7572
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.518684832490313e-05,
+      "loss": 1.6086,
+      "step": 7573
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.532647695949057e-05,
+      "loss": 1.7457,
+      "step": 7574
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.546615683510768e-05,
+      "loss": 3.5335,
+      "step": 7575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.560588766167265e-05,
+      "loss": 3.3315,
+      "step": 7576
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.574566914899782e-05,
+      "loss": 1.697,
+      "step": 7577
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.588550100679036e-05,
+      "loss": 5.4729,
+      "step": 7578
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.60253829446528e-05,
+      "loss": 2.1271,
+      "step": 7579
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.616531467208368e-05,
+      "loss": 2.1307,
+      "step": 7580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.630529589847797e-05,
+      "loss": 4.3793,
+      "step": 7581
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.644532633312833e-05,
+      "loss": 1.7718,
+      "step": 7582
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.658540568522477e-05,
+      "loss": 7.027,
+      "step": 7583
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.67255336638558e-05,
+      "loss": 5.9858,
+      "step": 7584
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.686570997800903e-05,
+      "loss": 1.3178,
+      "step": 7585
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.700593433657164e-05,
+      "loss": 3.6953,
+      "step": 7586
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.714620644833104e-05,
+      "loss": 2.813,
+      "step": 7587
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.728652602197544e-05,
+      "loss": 2.8516,
+      "step": 7588
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.742689276609454e-05,
+      "loss": 2.4475,
+      "step": 7589
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.756730638918006e-05,
+      "loss": 3.1334,
+      "step": 7590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.770776659962633e-05,
+      "loss": 3.8298,
+      "step": 7591
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.784827310573096e-05,
+      "loss": 3.4166,
+      "step": 7592
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.798882561569541e-05,
+      "loss": 3.1093,
+      "step": 7593
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.812942383762561e-05,
+      "loss": 2.5583,
+      "step": 7594
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.827006747953252e-05,
+      "loss": 2.7833,
+      "step": 7595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.841075624933283e-05,
+      "loss": 2.0695,
+      "step": 7596
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.855148985484946e-05,
+      "loss": 2.7684,
+      "step": 7597
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.869226800381223e-05,
+      "loss": 2.9088,
+      "step": 7598
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.883309040385832e-05,
+      "loss": 2.3687,
+      "step": 7599
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.897395676253345e-05,
+      "loss": 2.8818,
+      "step": 7600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.911486678729162e-05,
+      "loss": 2.7027,
+      "step": 7601
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.925582018549626e-05,
+      "loss": 3.0027,
+      "step": 7602
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.93968166644209e-05,
+      "loss": 3.3086,
+      "step": 7603
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.953785593124912e-05,
+      "loss": 3.5604,
+      "step": 7604
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.967893769307612e-05,
+      "loss": 2.9836,
+      "step": 7605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.982006165690868e-05,
+      "loss": 3.2076,
+      "step": 7606
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.996122752966597e-05,
+      "loss": 3.391,
+      "step": 7607
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.010243501818006e-05,
+      "loss": 3.516,
+      "step": 7608
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.024368382919669e-05,
+      "loss": 2.3564,
+      "step": 7609
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.038497366937568e-05,
+      "loss": 3.2252,
+      "step": 7610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.052630424529171e-05,
+      "loss": 4.4321,
+      "step": 7611
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.066767526343485e-05,
+      "loss": 3.4784,
+      "step": 7612
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.080908643021096e-05,
+      "loss": 3.8177,
+      "step": 7613
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.095053745194314e-05,
+      "loss": 2.2886,
+      "step": 7614
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.10920280348711e-05,
+      "loss": 1.9139,
+      "step": 7615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 8.12335578851526e-05,
+      "loss": 2.5648,
+      "step": 7616
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.137512670886389e-05,
+      "loss": 1.731,
+      "step": 7617
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.15167342120002e-05,
+      "loss": 3.3151,
+      "step": 7618
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.165838010047652e-05,
+      "loss": 3.1689,
+      "step": 7619
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.180006408012807e-05,
+      "loss": 3.6062,
+      "step": 7620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.194178585671097e-05,
+      "loss": 2.3503,
+      "step": 7621
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.208354513590286e-05,
+      "loss": 3.2923,
+      "step": 7622
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.222534162330348e-05,
+      "loss": 3.0461,
+      "step": 7623
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.23671750244353e-05,
+      "loss": 1.2581,
+      "step": 7624
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.250904504474414e-05,
+      "loss": 2.9254,
+      "step": 7625
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.265095138959977e-05,
+      "loss": 3.1763,
+      "step": 7626
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.27928937642965e-05,
+      "loss": 3.48,
+      "step": 7627
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.293487187405385e-05,
+      "loss": 3.216,
+      "step": 7628
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.30768854240171e-05,
+      "loss": 2.5144,
+      "step": 7629
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.321893411925774e-05,
+      "loss": 2.3066,
+      "step": 7630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.336101766477485e-05,
+      "loss": 2.2556,
+      "step": 7631
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.350313576549457e-05,
+      "loss": 2.4268,
+      "step": 7632
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.364528812627145e-05,
+      "loss": 3.207,
+      "step": 7633
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.378747445188891e-05,
+      "loss": 5.0218,
+      "step": 7634
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.392969444705999e-05,
+      "loss": 2.0527,
+      "step": 7635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.40719478164273e-05,
+      "loss": 1.9163,
+      "step": 7636
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.421423426456465e-05,
+      "loss": 2.9745,
+      "step": 7637
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.435655349597696e-05,
+      "loss": 2.0699,
+      "step": 7638
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.449890521510112e-05,
+      "loss": 1.8577,
+      "step": 7639
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.46412891263065e-05,
+      "loss": 1.8167,
+      "step": 7640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.478370493389567e-05,
+      "loss": 3.4506,
+      "step": 7641
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.492615234210488e-05,
+      "loss": 2.28,
+      "step": 7642
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.506863105510482e-05,
+      "loss": 2.1934,
+      "step": 7643
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.521114077700096e-05,
+      "loss": 4.2174,
+      "step": 7644
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.535368121183491e-05,
+      "loss": 2.3191,
+      "step": 7645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.549625206358396e-05,
+      "loss": 2.9375,
+      "step": 7646
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.563885303616244e-05,
+      "loss": 4.1788,
+      "step": 7647
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.578148383342205e-05,
+      "loss": 2.1125,
+      "step": 7648
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.592414415915266e-05,
+      "loss": 2.9978,
+      "step": 7649
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.606683371708273e-05,
+      "loss": 3.2788,
+      "step": 7650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.620955221088002e-05,
+      "loss": 3.1801,
+      "step": 7651
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.635229934415225e-05,
+      "loss": 1.7281,
+      "step": 7652
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.649507482044761e-05,
+      "loss": 3.0456,
+      "step": 7653
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.663787834325545e-05,
+      "loss": 3.2255,
+      "step": 7654
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.678070961600688e-05,
+      "loss": 1.3343,
+      "step": 7655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.692356834207537e-05,
+      "loss": 1.0574,
+      "step": 7656
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.706645422477736e-05,
+      "loss": 2.0064,
+      "step": 7657
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.720936696737292e-05,
+      "loss": 3.4544,
+      "step": 7658
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.735230627306634e-05,
+      "loss": 3.1752,
+      "step": 7659
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.749527184500673e-05,
+      "loss": 0.9107,
+      "step": 7660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.763826338628863e-05,
+      "loss": 3.6566,
+      "step": 7661
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.778128059995253e-05,
+      "loss": 4.0562,
+      "step": 7662
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.792432318898607e-05,
+      "loss": 3.8771,
+      "step": 7663
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.80673908563237e-05,
+      "loss": 2.4421,
+      "step": 7664
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.821048330484796e-05,
+      "loss": 2.31,
+      "step": 7665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.835360023738994e-05,
+      "loss": 2.7351,
+      "step": 7666
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.849674135673001e-05,
+      "loss": 3.8379,
+      "step": 7667
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.863990636559783e-05,
+      "loss": 4.1382,
+      "step": 7668
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.878309496667394e-05,
+      "loss": 2.8502,
+      "step": 7669
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.892630686258974e-05,
+      "loss": 2.3074,
+      "step": 7670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.906954175592823e-05,
+      "loss": 4.0399,
+      "step": 7671
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.921279934922467e-05,
+      "loss": 2.1816,
+      "step": 7672
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.935607934496715e-05,
+      "loss": 2.3449,
+      "step": 7673
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.949938144559725e-05,
+      "loss": 0.8652,
+      "step": 7674
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.964270535351067e-05,
+      "loss": 3.9267,
+      "step": 7675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.978605077105759e-05,
+      "loss": 2.5953,
+      "step": 7676
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 8.992941740054409e-05,
+      "loss": 3.0825,
+      "step": 7677
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.007280494423165e-05,
+      "loss": 2.7363,
+      "step": 7678
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.021621310433852e-05,
+      "loss": 1.8003,
+      "step": 7679
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.035964158304009e-05,
+      "loss": 3.5203,
+      "step": 7680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.050309008246957e-05,
+      "loss": 3.647,
+      "step": 7681
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.064655830471864e-05,
+      "loss": 3.6534,
+      "step": 7682
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.079004595183795e-05,
+      "loss": 2.8577,
+      "step": 7683
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.093355272583783e-05,
+      "loss": 3.2742,
+      "step": 7684
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.107707832868891e-05,
+      "loss": 4.6755,
+      "step": 7685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.12206224623227e-05,
+      "loss": 1.6138,
+      "step": 7686
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.136418482863223e-05,
+      "loss": 3.0709,
+      "step": 7687
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.150776512947268e-05,
+      "loss": 2.4346,
+      "step": 7688
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.165136306666193e-05,
+      "loss": 2.6045,
+      "step": 7689
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.179497834198132e-05,
+      "loss": 0.8651,
+      "step": 7690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.19386106571761e-05,
+      "loss": 5.431,
+      "step": 7691
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.208225971395617e-05,
+      "loss": 1.9063,
+      "step": 7692
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.222592521399649e-05,
+      "loss": 4.3414,
+      "step": 7693
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.236960685893841e-05,
+      "loss": 3.2712,
+      "step": 7694
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.251330435038916e-05,
+      "loss": 1.9969,
+      "step": 7695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.265701738992333e-05,
+      "loss": 2.5503,
+      "step": 7696
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.280074567908314e-05,
+      "loss": 2.8305,
+      "step": 7697
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.29444889193792e-05,
+      "loss": 4.5806,
+      "step": 7698
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.308824681229119e-05,
+      "loss": 4.6531,
+      "step": 7699
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.323201905926787e-05,
+      "loss": 2.3593,
+      "step": 7700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.33758053617287e-05,
+      "loss": 4.3543,
+      "step": 7701
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.351960542106381e-05,
+      "loss": 0.69,
+      "step": 7702
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.36634189386347e-05,
+      "loss": 4.9809,
+      "step": 7703
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.380724561577497e-05,
+      "loss": 4.1071,
+      "step": 7704
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.395108515379087e-05,
+      "loss": 2.0727,
+      "step": 7705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.409493725396195e-05,
+      "loss": 1.6771,
+      "step": 7706
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.42388016175415e-05,
+      "loss": 1.8365,
+      "step": 7707
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.438267794575782e-05,
+      "loss": 1.7456,
+      "step": 7708
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.452656593981392e-05,
+      "loss": 4.009,
+      "step": 7709
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.467046530088868e-05,
+      "loss": 3.1119,
+      "step": 7710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.481437573013742e-05,
+      "loss": 3.9691,
+      "step": 7711
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.495829692869246e-05,
+      "loss": 1.6407,
+      "step": 7712
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.510222859766376e-05,
+      "loss": 4.5051,
+      "step": 7713
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.524617043813952e-05,
+      "loss": 0.8467,
+      "step": 7714
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.539012215118684e-05,
+      "loss": 1.6744,
+      "step": 7715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.553408343785229e-05,
+      "loss": 3.507,
+      "step": 7716
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.567805399916258e-05,
+      "loss": 4.1043,
+      "step": 7717
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.582203353612514e-05,
+      "loss": 3.7841,
+      "step": 7718
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.596602174972879e-05,
+      "loss": 3.5984,
+      "step": 7719
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.611001834094428e-05,
+      "loss": 6.0942,
+      "step": 7720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.625402301072504e-05,
+      "loss": 3.3231,
+      "step": 7721
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.639803546000765e-05,
+      "loss": 1.7657,
+      "step": 7722
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.654205538971252e-05,
+      "loss": 2.2304,
+      "step": 7723
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.668608250074463e-05,
+      "loss": 1.7598,
+      "step": 7724
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.683011649399377e-05,
+      "loss": 4.2121,
+      "step": 7725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.6974157070336e-05,
+      "loss": 2.0959,
+      "step": 7726
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.71182039306332e-05,
+      "loss": 2.5158,
+      "step": 7727
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.726225677573433e-05,
+      "loss": 3.4428,
+      "step": 7728
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.740631530647597e-05,
+      "loss": 4.8221,
+      "step": 7729
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.755037922368284e-05,
+      "loss": 1.2524,
+      "step": 7730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.769444822816868e-05,
+      "loss": 2.7174,
+      "step": 7731
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.783852202073617e-05,
+      "loss": 2.3986,
+      "step": 7732
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.798260030217853e-05,
+      "loss": 3.2976,
+      "step": 7733
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.812668277327947e-05,
+      "loss": 1.9456,
+      "step": 7734
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.827076913481402e-05,
+      "loss": 4.4385,
+      "step": 7735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.841485908754916e-05,
+      "loss": 4.9741,
+      "step": 7736
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.855895233224437e-05,
+      "loss": 3.8754,
+      "step": 7737
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.87030485696523e-05,
+      "loss": 4.3715,
+      "step": 7738
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.88471475005192e-05,
+      "loss": 2.4167,
+      "step": 7739
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.89912488255863e-05,
+      "loss": 3.1939,
+      "step": 7740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.913535224558928e-05,
+      "loss": 1.6198,
+      "step": 7741
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.927945746125968e-05,
+      "loss": 2.2559,
+      "step": 7742
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.942356417332529e-05,
+      "loss": 1.5387,
+      "step": 7743
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.956767208251078e-05,
+      "loss": 4.0999,
+      "step": 7744
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.971178088953837e-05,
+      "loss": 1.1703,
+      "step": 7745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.98558902951284e-05,
+      "loss": 3.2932,
+      "step": 7746
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.999999999999996e-05,
+      "loss": 3.583,
+      "step": 7747
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010014410970487151,
+      "loss": 0.6894,
+      "step": 7748
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010028821911046156,
+      "loss": 2.7168,
+      "step": 7749
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010043232791748913,
+      "loss": 3.7921,
+      "step": 7750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010057643582667464,
+      "loss": 4.3406,
+      "step": 7751
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010072054253874024,
+      "loss": 3.6586,
+      "step": 7752
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010086464775441065,
+      "loss": 2.158,
+      "step": 7753
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010100875117441363,
+      "loss": 3.1089,
+      "step": 7754
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010115285249948073,
+      "loss": 1.9651,
+      "step": 7755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010129695143034763,
+      "loss": 3.5016,
+      "step": 7756
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010144104766775555,
+      "loss": 2.9668,
+      "step": 7757
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010158514091245077,
+      "loss": 3.7203,
+      "step": 7758
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010172923086518588,
+      "loss": 3.0094,
+      "step": 7759
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010187331722672047,
+      "loss": 3.2645,
+      "step": 7760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001020173996978214,
+      "loss": 3.7346,
+      "step": 7761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010216147797926374,
+      "loss": 3.7483,
+      "step": 7762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010230555177183123,
+      "loss": 2.7063,
+      "step": 7763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001024496207763171,
+      "loss": 2.016,
+      "step": 7764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010259368469352398,
+      "loss": 3.1696,
+      "step": 7765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001027377432242656,
+      "loss": 1.1128,
+      "step": 7766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010288179606936672,
+      "loss": 1.6391,
+      "step": 7767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001030258429296639,
+      "loss": 2.5345,
+      "step": 7768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010316988350600616,
+      "loss": 2.057,
+      "step": 7769
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001033139174992553,
+      "loss": 0.3561,
+      "step": 7770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010345794461028739,
+      "loss": 2.842,
+      "step": 7771
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001036019645399923,
+      "loss": 3.2338,
+      "step": 7772
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010374597698927489,
+      "loss": 3.2023,
+      "step": 7773
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010388998165905563,
+      "loss": 1.6668,
+      "step": 7774
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010403397825027115,
+      "loss": 2.5867,
+      "step": 7775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010417796646387479,
+      "loss": 5.2234,
+      "step": 7776
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010432194600083735,
+      "loss": 4.6046,
+      "step": 7777
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010446591656214764,
+      "loss": 1.568,
+      "step": 7778
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010460987784881309,
+      "loss": 1.6603,
+      "step": 7779
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010475382956186041,
+      "loss": 2.0826,
+      "step": 7780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010489777140233617,
+      "loss": 3.6703,
+      "step": 7781
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010504170307130747,
+      "loss": 4.0345,
+      "step": 7782
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010518562426986252,
+      "loss": 2.5964,
+      "step": 7783
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010532953469911125,
+      "loss": 3.2227,
+      "step": 7784
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010547343406018601,
+      "loss": 2.8477,
+      "step": 7785
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010561732205424211,
+      "loss": 3.4564,
+      "step": 7786
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010576119838245844,
+      "loss": 1.8244,
+      "step": 7787
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010590506274603798,
+      "loss": 2.8619,
+      "step": 7788
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010604891484620904,
+      "loss": 2.4801,
+      "step": 7789
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010619275438422496,
+      "loss": 4.1306,
+      "step": 7790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010633658106136521,
+      "loss": 3.1461,
+      "step": 7791
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010648039457893613,
+      "loss": 2.5918,
+      "step": 7792
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010662419463827123,
+      "loss": 3.1441,
+      "step": 7793
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010676798094073206,
+      "loss": 1.8306,
+      "step": 7794
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010691175318770874,
+      "loss": 2.7765,
+      "step": 7795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010705551108062073,
+      "loss": 2.442,
+      "step": 7796
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010719925432091677,
+      "loss": 2.5809,
+      "step": 7797
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001073429826100766,
+      "loss": 0.754,
+      "step": 7798
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010748669564961076,
+      "loss": 2.0762,
+      "step": 7799
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00010763039314106152,
+      "loss": 3.7349,
+      "step": 7800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010777407478600343,
+      "loss": 5.3274,
+      "step": 7801
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010791774028604376,
+      "loss": 3.357,
+      "step": 7802
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010806138934282384,
+      "loss": 3.403,
+      "step": 7803
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001082050216580186,
+      "loss": 4.823,
+      "step": 7804
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000108348636933338,
+      "loss": 3.5865,
+      "step": 7805
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010849223487052725,
+      "loss": 1.9961,
+      "step": 7806
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001086358151713677,
+      "loss": 3.5669,
+      "step": 7807
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010877937753767722,
+      "loss": 2.9332,
+      "step": 7808
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010892292167131102,
+      "loss": 3.3196,
+      "step": 7809
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010906644727416211,
+      "loss": 1.1217,
+      "step": 7810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010920995404816198,
+      "loss": 1.943,
+      "step": 7811
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010935344169528128,
+      "loss": 0.2274,
+      "step": 7812
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010949690991753034,
+      "loss": 0.7301,
+      "step": 7813
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010964035841695984,
+      "loss": 1.6495,
+      "step": 7814
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001097837868956614,
+      "loss": 3.6615,
+      "step": 7815
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00010992719505576827,
+      "loss": 3.3569,
+      "step": 7816
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011007058259945584,
+      "loss": 1.5988,
+      "step": 7817
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011021394922894235,
+      "loss": 1.4533,
+      "step": 7818
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011035729464648926,
+      "loss": 1.2649,
+      "step": 7819
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011050061855440267,
+      "loss": 4.2092,
+      "step": 7820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011064392065503278,
+      "loss": 3.2559,
+      "step": 7821
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011078720065077526,
+      "loss": 3.6083,
+      "step": 7822
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011093045824407169,
+      "loss": 2.8962,
+      "step": 7823
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011107369313741018,
+      "loss": 3.2075,
+      "step": 7824
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011121690503332598,
+      "loss": 4.5319,
+      "step": 7825
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001113600936344021,
+      "loss": 3.1372,
+      "step": 7826
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011150325864326992,
+      "loss": 4.5115,
+      "step": 7827
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011164639976260997,
+      "loss": 2.0654,
+      "step": 7828
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011178951669515197,
+      "loss": 2.343,
+      "step": 7829
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011193260914367623,
+      "loss": 2.8624,
+      "step": 7830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011207567681101385,
+      "loss": 1.3986,
+      "step": 7831
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001122187194000474,
+      "loss": 3.9597,
+      "step": 7832
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001123617366137113,
+      "loss": 2.9485,
+      "step": 7833
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001125047281549932,
+      "loss": 2.4981,
+      "step": 7834
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011264769372693357,
+      "loss": 4.5883,
+      "step": 7835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000112790633032627,
+      "loss": 1.9371,
+      "step": 7836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011293354577522257,
+      "loss": 1.3582,
+      "step": 7837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011307643165792456,
+      "loss": 3.2578,
+      "step": 7838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011321929038399304,
+      "loss": 4.0689,
+      "step": 7839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011336212165674447,
+      "loss": 3.5368,
+      "step": 7840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011350492517955232,
+      "loss": 4.7194,
+      "step": 7841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011364770065584766,
+      "loss": 4.7218,
+      "step": 7842
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011379044778911989,
+      "loss": 5.2853,
+      "step": 7843
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011393316628291719,
+      "loss": 3.7505,
+      "step": 7844
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011407585584084726,
+      "loss": 3.0125,
+      "step": 7845
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011421851616657786,
+      "loss": 4.1637,
+      "step": 7846
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011436114696383749,
+      "loss": 3.3314,
+      "step": 7847
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011450374793641596,
+      "loss": 3.3954,
+      "step": 7848
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011464631878816501,
+      "loss": 2.9301,
+      "step": 7849
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011478885922299896,
+      "loss": 2.8895,
+      "step": 7850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011493136894489512,
+      "loss": 3.9485,
+      "step": 7851
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011507384765789503,
+      "loss": 1.541,
+      "step": 7852
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011521629506610426,
+      "loss": 3.2542,
+      "step": 7853
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011535871087369341,
+      "loss": 2.1284,
+      "step": 7854
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001155010947848988,
+      "loss": 3.4812,
+      "step": 7855
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011564344650402297,
+      "loss": 2.5577,
+      "step": 7856
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001157857657354353,
+      "loss": 1.5778,
+      "step": 7857
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011592805218357263,
+      "loss": 3.5092,
+      "step": 7858
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011607030555293994,
+      "loss": 3.882,
+      "step": 7859
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011621252554811102,
+      "loss": 0.5867,
+      "step": 7860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011635471187372848,
+      "loss": 2.7873,
+      "step": 7861
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011649686423450536,
+      "loss": 4.0638,
+      "step": 7862
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011663898233522508,
+      "loss": 2.0081,
+      "step": 7863
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011678106588074218,
+      "loss": 3.3006,
+      "step": 7864
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011692311457598283,
+      "loss": 3.6409,
+      "step": 7865
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011706512812594608,
+      "loss": 2.6624,
+      "step": 7866
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011720710623570341,
+      "loss": 3.4539,
+      "step": 7867
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011734904861040015,
+      "loss": 2.4336,
+      "step": 7868
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011749095495525577,
+      "loss": 3.7256,
+      "step": 7869
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011763282497556463,
+      "loss": 4.4876,
+      "step": 7870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011777465837669645,
+      "loss": 3.0964,
+      "step": 7871
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011791645486409706,
+      "loss": 1.535,
+      "step": 7872
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011805821414328894,
+      "loss": 3.0979,
+      "step": 7873
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011819993591987186,
+      "loss": 3.7003,
+      "step": 7874
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001183416198995234,
+      "loss": 0.3626,
+      "step": 7875
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011848326578799973,
+      "loss": 4.5361,
+      "step": 7876
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011862487329113606,
+      "loss": 2.5889,
+      "step": 7877
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011876644211484733,
+      "loss": 5.7218,
+      "step": 7878
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011890797196512884,
+      "loss": 1.9083,
+      "step": 7879
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011904946254805677,
+      "loss": 2.4307,
+      "step": 7880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011919091356978895,
+      "loss": 3.4835,
+      "step": 7881
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011933232473656509,
+      "loss": 2.5404,
+      "step": 7882
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011947369575470821,
+      "loss": 3.5596,
+      "step": 7883
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011961502633062424,
+      "loss": 2.4297,
+      "step": 7884
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011975631617080323,
+      "loss": 2.2876,
+      "step": 7885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00011989756498181985,
+      "loss": 1.6252,
+      "step": 7886
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012003877247033397,
+      "loss": 3.6251,
+      "step": 7887
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012017993834309125,
+      "loss": 4.4167,
+      "step": 7888
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012032106230692381,
+      "loss": 3.1325,
+      "step": 7889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001204621440687508,
+      "loss": 2.8951,
+      "step": 7890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012060318333557903,
+      "loss": 5.487,
+      "step": 7891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012074417981450368,
+      "loss": 3.9746,
+      "step": 7892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012088513321270831,
+      "loss": 3.7401,
+      "step": 7893
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012102604323746647,
+      "loss": 3.7032,
+      "step": 7894
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012116690959614161,
+      "loss": 3.5543,
+      "step": 7895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001213077319961877,
+      "loss": 3.1739,
+      "step": 7896
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012144851014515048,
+      "loss": 3.3286,
+      "step": 7897
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012158924375066709,
+      "loss": 2.2972,
+      "step": 7898
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001217299325204674,
+      "loss": 4.522,
+      "step": 7899
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012187057616237432,
+      "loss": 0.9613,
+      "step": 7900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012201117438430452,
+      "loss": 3.4938,
+      "step": 7901
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012215172689426898,
+      "loss": 4.0447,
+      "step": 7902
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001222922334003736,
+      "loss": 0.7112,
+      "step": 7903
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012243269361081986,
+      "loss": 3.0724,
+      "step": 7904
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012257310723390538,
+      "loss": 1.847,
+      "step": 7905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001227134739780245,
+      "loss": 5.4967,
+      "step": 7906
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012285379355166887,
+      "loss": 4.14,
+      "step": 7907
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012299406566342828,
+      "loss": 4.9027,
+      "step": 7908
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012313429002199088,
+      "loss": 3.0144,
+      "step": 7909
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001232744663361441,
+      "loss": 2.3348,
+      "step": 7910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012341459431477515,
+      "loss": 4.603,
+      "step": 7911
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012355467366687158,
+      "loss": 4.5044,
+      "step": 7912
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012369470410152194,
+      "loss": 0.2889,
+      "step": 7913
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012383468532791624,
+      "loss": 2.0559,
+      "step": 7914
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012397461705534712,
+      "loss": 3.3636,
+      "step": 7915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012411449899320958,
+      "loss": 1.2009,
+      "step": 7916
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001242543308510021,
+      "loss": 4.4006,
+      "step": 7917
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012439411233832728,
+      "loss": 3.3414,
+      "step": 7918
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012453384316489226,
+      "loss": 3.2133,
+      "step": 7919
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012467352304050937,
+      "loss": 3.3153,
+      "step": 7920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012481315167509678,
+      "loss": 1.7168,
+      "step": 7921
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012495272877867915,
+      "loss": 2.2199,
+      "step": 7922
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000125092254061388,
+      "loss": 4.3224,
+      "step": 7923
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012523172723346277,
+      "loss": 3.0858,
+      "step": 7924
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012537114800525054,
+      "loss": 3.8113,
+      "step": 7925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012551051608720778,
+      "loss": 1.399,
+      "step": 7926
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012564983118990017,
+      "loss": 1.749,
+      "step": 7927
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012578909302400325,
+      "loss": 3.2416,
+      "step": 7928
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012592830130030372,
+      "loss": 2.4278,
+      "step": 7929
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012606745572969898,
+      "loss": 1.6284,
+      "step": 7930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012620655602319845,
+      "loss": 3.33,
+      "step": 7931
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012634560189192387,
+      "loss": 4.0659,
+      "step": 7932
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012648459304711023,
+      "loss": 2.8132,
+      "step": 7933
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001266235292001059,
+      "loss": 2.2501,
+      "step": 7934
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012676241006237361,
+      "loss": 1.01,
+      "step": 7935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001269012353454909,
+      "loss": 2.0469,
+      "step": 7936
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012704000476115076,
+      "loss": 0.4432,
+      "step": 7937
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012717871802116215,
+      "loss": 5.0875,
+      "step": 7938
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012731737483745066,
+      "loss": 1.127,
+      "step": 7939
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001274559749220591,
+      "loss": 1.7048,
+      "step": 7940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012759451798714815,
+      "loss": 2.2318,
+      "step": 7941
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012773300374499687,
+      "loss": 0.2686,
+      "step": 7942
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012787143190800328,
+      "loss": 3.5425,
+      "step": 7943
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012800980218868513,
+      "loss": 3.9451,
+      "step": 7944
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012814811429968023,
+      "loss": 1.9436,
+      "step": 7945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012828636795374717,
+      "loss": 4.6147,
+      "step": 7946
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012842456286376634,
+      "loss": 2.6104,
+      "step": 7947
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001285626987427397,
+      "loss": 1.8902,
+      "step": 7948
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001287007753037919,
+      "loss": 2.4936,
+      "step": 7949
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012883879226017087,
+      "loss": 2.378,
+      "step": 7950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001289767493252482,
+      "loss": 0.8126,
+      "step": 7951
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012911464621252,
+      "loss": 6.0579,
+      "step": 7952
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001292524826356072,
+      "loss": 2.3643,
+      "step": 7953
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012939025830825646,
+      "loss": 2.6676,
+      "step": 7954
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012952797294434049,
+      "loss": 0.8251,
+      "step": 7955
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000129665626257859,
+      "loss": 0.7294,
+      "step": 7956
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012980321796293847,
+      "loss": 2.5874,
+      "step": 7957
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012994074777383387,
+      "loss": 4.1799,
+      "step": 7958
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013007821540492843,
+      "loss": 2.8992,
+      "step": 7959
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013021562057073492,
+      "loss": 4.7719,
+      "step": 7960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001303529629858954,
+      "loss": 3.8743,
+      "step": 7961
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013049024236518237,
+      "loss": 2.3229,
+      "step": 7962
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013062745842349927,
+      "loss": 4.7848,
+      "step": 7963
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013076461087588106,
+      "loss": 3.4507,
+      "step": 7964
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001309016994374947,
+      "loss": 2.2285,
+      "step": 7965
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013103872382363998,
+      "loss": 5.5696,
+      "step": 7966
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013117568374974986,
+      "loss": 4.1644,
+      "step": 7967
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013131257893139118,
+      "loss": 1.7768,
+      "step": 7968
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013144940908426528,
+      "loss": 2.4751,
+      "step": 7969
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013158617392420853,
+      "loss": 1.0179,
+      "step": 7970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013172287316719295,
+      "loss": 2.5232,
+      "step": 7971
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013185950652932671,
+      "loss": 1.7943,
+      "step": 7972
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013199607372685497,
+      "loss": 2.2527,
+      "step": 7973
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001321325744761601,
+      "loss": 2.9424,
+      "step": 7974
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013226900849376262,
+      "loss": 1.8183,
+      "step": 7975
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013240537549632155,
+      "loss": 2.5446,
+      "step": 7976
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013254167520063495,
+      "loss": 0.5743,
+      "step": 7977
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013267790732364115,
+      "loss": 2.1386,
+      "step": 7978
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013281407158241828,
+      "loss": 4.9306,
+      "step": 7979
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013295016769418564,
+      "loss": 1.8375,
+      "step": 7980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013308619537630404,
+      "loss": 1.9113,
+      "step": 7981
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013322215434627635,
+      "loss": 1.7362,
+      "step": 7982
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013335804432174818,
+      "loss": 3.4001,
+      "step": 7983
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013349386502050847,
+      "loss": 3.6426,
+      "step": 7984
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013362961616048992,
+      "loss": 3.0522,
+      "step": 7985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013376529745976983,
+      "loss": 0.3987,
+      "step": 7986
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001339009086365704,
+      "loss": 3.9774,
+      "step": 7987
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013403644940925967,
+      "loss": 2.3494,
+      "step": 7988
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013417191949635147,
+      "loss": 4.2225,
+      "step": 7989
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013430731861650684,
+      "loss": 3.2612,
+      "step": 7990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013444264648853387,
+      "loss": 3.441,
+      "step": 7991
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001345779028313892,
+      "loss": 1.8559,
+      "step": 7992
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001347130873641774,
+      "loss": 1.1901,
+      "step": 7993
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013484819980615247,
+      "loss": 3.909,
+      "step": 7994
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013498323987671797,
+      "loss": 3.465,
+      "step": 7995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013511820729542788,
+      "loss": 4.1555,
+      "step": 7996
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013525310178198705,
+      "loss": 1.9349,
+      "step": 7997
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001353879230562517,
+      "loss": 4.379,
+      "step": 7998
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013552267083823024,
+      "loss": 2.7051,
+      "step": 7999
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013565734484808357,
+      "loss": 4.0125,
+      "step": 8000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013579194480612587,
+      "loss": 2.2651,
+      "step": 8001
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001359264704328251,
+      "loss": 2.935,
+      "step": 8002
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001360609214488036,
+      "loss": 3.1222,
+      "step": 8003
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013619529757483856,
+      "loss": 4.7038,
+      "step": 8004
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013632959853186283,
+      "loss": 3.0703,
+      "step": 8005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013646382404096532,
+      "loss": 1.5461,
+      "step": 8006
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001365979738233916,
+      "loss": 4.4623,
+      "step": 8007
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013673204760054458,
+      "loss": 1.4144,
+      "step": 8008
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013686604509398475,
+      "loss": 2.8053,
+      "step": 8009
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013699996602543156,
+      "loss": 1.4284,
+      "step": 8010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000137133810116763,
+      "loss": 2.2658,
+      "step": 8011
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013726757709001678,
+      "loss": 1.0809,
+      "step": 8012
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013740126666739072,
+      "loss": 2.6491,
+      "step": 8013
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001375348785712435,
+      "loss": 4.627,
+      "step": 8014
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013766841252409493,
+      "loss": 1.8719,
+      "step": 8015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013780186824862684,
+      "loss": 4.2078,
+      "step": 8016
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013793524546768348,
+      "loss": 2.736,
+      "step": 8017
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013806854390427216,
+      "loss": 2.4987,
+      "step": 8018
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013820176328156373,
+      "loss": 2.2169,
+      "step": 8019
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001383349033228935,
+      "loss": 4.0949,
+      "step": 8020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013846796375176093,
+      "loss": 3.5954,
+      "step": 8021
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013860094429183145,
+      "loss": 2.5087,
+      "step": 8022
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013873384466693594,
+      "loss": 2.5072,
+      "step": 8023
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001388666646010723,
+      "loss": 2.314,
+      "step": 8024
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013899940381840504,
+      "loss": 2.7164,
+      "step": 8025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001391320620432664,
+      "loss": 4.166,
+      "step": 8026
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013926463900015685,
+      "loss": 1.757,
+      "step": 8027
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013939713441374566,
+      "loss": 3.1558,
+      "step": 8028
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013952954800887135,
+      "loss": 0.9031,
+      "step": 8029
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001396618795105425,
+      "loss": 4.4185,
+      "step": 8030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013979412864393808,
+      "loss": 3.0466,
+      "step": 8031
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00013992629513440817,
+      "loss": 2.7514,
+      "step": 8032
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014005837870747438,
+      "loss": 2.4376,
+      "step": 8033
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014019037908883066,
+      "loss": 2.4312,
+      "step": 8034
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014032229600434364,
+      "loss": 3.571,
+      "step": 8035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014045412918005333,
+      "loss": 2.7655,
+      "step": 8036
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014058587834217358,
+      "loss": 5.4659,
+      "step": 8037
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001407175432170928,
+      "loss": 1.4893,
+      "step": 8038
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014084912353137444,
+      "loss": 3.3871,
+      "step": 8039
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014098061901175732,
+      "loss": 4.773,
+      "step": 8040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000141112029385157,
+      "loss": 2.6699,
+      "step": 8041
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014124335437866526,
+      "loss": 4.0334,
+      "step": 8042
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014137459371955138,
+      "loss": 2.4752,
+      "step": 8043
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014150574713526253,
+      "loss": 4.5404,
+      "step": 8044
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001416368143534243,
+      "loss": 3.7051,
+      "step": 8045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014176779510184127,
+      "loss": 1.8497,
+      "step": 8046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014189868910849767,
+      "loss": 2.2597,
+      "step": 8047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001420294961015578,
+      "loss": 3.0712,
+      "step": 8048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014216021580936672,
+      "loss": 2.377,
+      "step": 8049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014229084796045068,
+      "loss": 0.3958,
+      "step": 8050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014242139228351792,
+      "loss": 3.3743,
+      "step": 8051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014255184850745908,
+      "loss": 3.1855,
+      "step": 8052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001426822163613473,
+      "loss": 3.4192,
+      "step": 8053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001428124955744397,
+      "loss": 2.0162,
+      "step": 8054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014294268587617762,
+      "loss": 4.4839,
+      "step": 8055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014307278699618673,
+      "loss": 1.348,
+      "step": 8056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014320279866427793,
+      "loss": 3.902,
+      "step": 8057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014333272061044797,
+      "loss": 3.2189,
+      "step": 8058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014346255256487996,
+      "loss": 3.6313,
+      "step": 8059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014359229425794386,
+      "loss": 2.6834,
+      "step": 8060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014372194542019702,
+      "loss": 2.0175,
+      "step": 8061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014385150578238488,
+      "loss": 2.8784,
+      "step": 8062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014398097507544146,
+      "loss": 2.7375,
+      "step": 8063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014411035303048988,
+      "loss": 3.4012,
+      "step": 8064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014423963937884292,
+      "loss": 1.3024,
+      "step": 8065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014436883385200364,
+      "loss": 3.7155,
+      "step": 8066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001444979361816659,
+      "loss": 3.5582,
+      "step": 8067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014462694609971495,
+      "loss": 3.5007,
+      "step": 8068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014475586333822786,
+      "loss": 5.0101,
+      "step": 8069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001448846876294743,
+      "loss": 3.3785,
+      "step": 8070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014501341870591682,
+      "loss": 3.236,
+      "step": 8071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014514205630021158,
+      "loss": 1.6527,
+      "step": 8072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001452706001452092,
+      "loss": 1.7076,
+      "step": 8073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014539904997395458,
+      "loss": 2.1844,
+      "step": 8074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000145527405519688,
+      "loss": 2.4728,
+      "step": 8075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014565566651584565,
+      "loss": 0.8194,
+      "step": 8076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014578383269605994,
+      "loss": 2.5612,
+      "step": 8077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014591190379416023,
+      "loss": 2.9996,
+      "step": 8078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001460398795441734,
+      "loss": 6.418,
+      "step": 8079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014616775968032426,
+      "loss": 3.6521,
+      "step": 8080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014629554393703626,
+      "loss": 0.6973,
+      "step": 8081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014642323204893195,
+      "loss": 3.3294,
+      "step": 8082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001465508237508335,
+      "loss": 2.0262,
+      "step": 8083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001466783187777635,
+      "loss": 5.9623,
+      "step": 8084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014680571686494492,
+      "loss": 2.6369,
+      "step": 8085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014693301774780217,
+      "loss": 4.6559,
+      "step": 8086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000147060221161962,
+      "loss": 4.4171,
+      "step": 8087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001471873268432531,
+      "loss": 2.0178,
+      "step": 8088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014731433452770716,
+      "loss": 1.0998,
+      "step": 8089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014744124395155958,
+      "loss": 4.2443,
+      "step": 8090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014756805485124965,
+      "loss": 4.0181,
+      "step": 8091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014769476696342135,
+      "loss": 2.0697,
+      "step": 8092
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014782138002492382,
+      "loss": 3.1174,
+      "step": 8093
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014794789377281192,
+      "loss": 2.4336,
+      "step": 8094
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014807430794434664,
+      "loss": 1.8935,
+      "step": 8095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000148200622276996,
+      "loss": 4.0498,
+      "step": 8096
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001483268365084351,
+      "loss": 0.6057,
+      "step": 8097
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014845295037654708,
+      "loss": 4.3313,
+      "step": 8098
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001485789636194235,
+      "loss": 2.2608,
+      "step": 8099
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014870487597536485,
+      "loss": 3.4602,
+      "step": 8100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001488306871828812,
+      "loss": 1.8449,
+      "step": 8101
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014895639698069262,
+      "loss": 4.658,
+      "step": 8102
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014908200510772965,
+      "loss": 2.7036,
+      "step": 8103
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014920751130313443,
+      "loss": 1.6265,
+      "step": 8104
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014933291530626032,
+      "loss": 3.4385,
+      "step": 8105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014945821685667313,
+      "loss": 1.6799,
+      "step": 8106
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014958341569415136,
+      "loss": 2.5646,
+      "step": 8107
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014970851155868687,
+      "loss": 2.6035,
+      "step": 8108
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014983350419048533,
+      "loss": 3.9923,
+      "step": 8109
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001499583933299668,
+      "loss": 1.6867,
+      "step": 8110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015008317871776633,
+      "loss": 2.1556,
+      "step": 8111
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015020786009473436,
+      "loss": 2.0285,
+      "step": 8112
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001503324372019374,
+      "loss": 3.6456,
+      "step": 8113
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015045690978065843,
+      "loss": 3.1152,
+      "step": 8114
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015058127757239763,
+      "loss": 2.2575,
+      "step": 8115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015070554031887282,
+      "loss": 1.9456,
+      "step": 8116
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015082969776201942,
+      "loss": 3.7855,
+      "step": 8117
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015095374964399247,
+      "loss": 5.4644,
+      "step": 8118
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001510776957071656,
+      "loss": 2.8018,
+      "step": 8119
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001512015356941324,
+      "loss": 4.7981,
+      "step": 8120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015132526934770666,
+      "loss": 4.3755,
+      "step": 8121
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015144889641092307,
+      "loss": 2.585,
+      "step": 8122
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001515724166270377,
+      "loss": 1.4804,
+      "step": 8123
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015169582973952853,
+      "loss": 2.6969,
+      "step": 8124
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001518191354920958,
+      "loss": 3.6115,
+      "step": 8125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001519423336286629,
+      "loss": 3.099,
+      "step": 8126
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015206542389337665,
+      "loss": 2.7649,
+      "step": 8127
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015218840603060794,
+      "loss": 1.7958,
+      "step": 8128
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015231127978495207,
+      "loss": 2.5735,
+      "step": 8129
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015243404490122958,
+      "loss": 2.845,
+      "step": 8130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001525567011244866,
+      "loss": 3.7683,
+      "step": 8131
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015267924819999527,
+      "loss": 3.9772,
+      "step": 8132
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015280168587325464,
+      "loss": 3.4537,
+      "step": 8133
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015292401388999072,
+      "loss": 2.9224,
+      "step": 8134
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015304623199615723,
+      "loss": 4.1726,
+      "step": 8135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015316833993793665,
+      "loss": 2.2683,
+      "step": 8136
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015329033746173964,
+      "loss": 2.0241,
+      "step": 8137
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015341222431420644,
+      "loss": 1.1929,
+      "step": 8138
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001535340002422071,
+      "loss": 1.3135,
+      "step": 8139
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000153655664992842,
+      "loss": 2.7381,
+      "step": 8140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015377721831344245,
+      "loss": 4.3395,
+      "step": 8141
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015389865995157117,
+      "loss": 2.373,
+      "step": 8142
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001540199896550228,
+      "loss": 3.5634,
+      "step": 8143
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015414120717182448,
+      "loss": 3.6221,
+      "step": 8144
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015426231225023623,
+      "loss": 4.0304,
+      "step": 8145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001543833046387517,
+      "loss": 1.9497,
+      "step": 8146
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015450418408609852,
+      "loss": 3.2263,
+      "step": 8147
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015462495034123901,
+      "loss": 3.0069,
+      "step": 8148
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015474560315337002,
+      "loss": 1.9856,
+      "step": 8149
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015486614227192477,
+      "loss": 3.0168,
+      "step": 8150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015498656744657215,
+      "loss": 1.7575,
+      "step": 8151
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015510687842721773,
+      "loss": 3.8912,
+      "step": 8152
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015522707496400424,
+      "loss": 4.0057,
+      "step": 8153
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015534715680731216,
+      "loss": 2.625,
+      "step": 8154
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015546712370776005,
+      "loss": 4.0864,
+      "step": 8155
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015558697541620526,
+      "loss": 4.8473,
+      "step": 8156
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015570671168374438,
+      "loss": 2.0821,
+      "step": 8157
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015582633226171363,
+      "loss": 2.7024,
+      "step": 8158
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015594583690168966,
+      "loss": 2.535,
+      "step": 8159
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015606522535548969,
+      "loss": 2.0861,
+      "step": 8160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015618449737517245,
+      "loss": 3.4707,
+      "step": 8161
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001563036527130383,
+      "loss": 1.1302,
+      "step": 8162
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015642269112162998,
+      "loss": 3.4577,
+      "step": 8163
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001565416123537331,
+      "loss": 5.5968,
+      "step": 8164
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015666041616237662,
+      "loss": 2.406,
+      "step": 8165
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015677910230083315,
+      "loss": 3.0299,
+      "step": 8166
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00015689767052262012,
+      "loss": 1.3851,
+      "step": 8167
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015701612058149949,
+      "loss": 2.7219,
+      "step": 8168
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015713445223147865,
+      "loss": 3.0212,
+      "step": 8169
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015725266522681098,
+      "loss": 3.9994,
+      "step": 8170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015737075932199628,
+      "loss": 3.3309,
+      "step": 8171
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001574887342717812,
+      "loss": 2.3476,
+      "step": 8172
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015760658983115985,
+      "loss": 4.1656,
+      "step": 8173
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015772432575537437,
+      "loss": 4.7028,
+      "step": 8174
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015784194179991527,
+      "loss": 2.3729,
+      "step": 8175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015795943772052204,
+      "loss": 2.7491,
+      "step": 8176
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015807681327318363,
+      "loss": 1.4784,
+      "step": 8177
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000158194068214139,
+      "loss": 2.531,
+      "step": 8178
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015831120229987757,
+      "loss": 5.2443,
+      "step": 8179
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015842821528713974,
+      "loss": 2.1386,
+      "step": 8180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001585451069329174,
+      "loss": 3.2004,
+      "step": 8181
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015866187699445442,
+      "loss": 2.9398,
+      "step": 8182
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001587785252292473,
+      "loss": 2.3576,
+      "step": 8183
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015889505139504533,
+      "loss": 2.7453,
+      "step": 8184
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001590114552498515,
+      "loss": 1.4334,
+      "step": 8185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015912773655192268,
+      "loss": 5.3389,
+      "step": 8186
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015924389505977033,
+      "loss": 5.0255,
+      "step": 8187
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001593599305321609,
+      "loss": 3.2959,
+      "step": 8188
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001594758427281164,
+      "loss": 1.1327,
+      "step": 8189
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001595916314069147,
+      "loss": 2.4482,
+      "step": 8190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015970729632809044,
+      "loss": 2.4103,
+      "step": 8191
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000159822837251435,
+      "loss": 3.4848,
+      "step": 8192
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015993825393699747,
+      "loss": 3.8442,
+      "step": 8193
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016005354614508487,
+      "loss": 2.5448,
+      "step": 8194
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001601687136362628,
+      "loss": 1.1484,
+      "step": 8195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016028375617135573,
+      "loss": 3.3643,
+      "step": 8196
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001603986735114478,
+      "loss": 4.2619,
+      "step": 8197
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016051346541788285,
+      "loss": 3.5509,
+      "step": 8198
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001606281316522658,
+      "loss": 1.4426,
+      "step": 8199
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000160742671976462,
+      "loss": 3.1216,
+      "step": 8200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016085708615259852,
+      "loss": 3.9637,
+      "step": 8201
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016097137394306435,
+      "loss": 2.1984,
+      "step": 8202
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016108553511051096,
+      "loss": 1.7584,
+      "step": 8203
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016119956941785285,
+      "loss": 3.0154,
+      "step": 8204
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016131347662826784,
+      "loss": 2.1313,
+      "step": 8205
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016142725650519787,
+      "loss": 3.0588,
+      "step": 8206
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016154090881234922,
+      "loss": 2.0611,
+      "step": 8207
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001616544333136931,
+      "loss": 2.5401,
+      "step": 8208
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016176782977346622,
+      "loss": 2.8365,
+      "step": 8209
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016188109795617104,
+      "loss": 2.6884,
+      "step": 8210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016199423762657664,
+      "loss": 4.4944,
+      "step": 8211
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016210724854971882,
+      "loss": 1.8818,
+      "step": 8212
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016222013049090083,
+      "loss": 1.32,
+      "step": 8213
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016233288321569375,
+      "loss": 2.1733,
+      "step": 8214
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016244550648993703,
+      "loss": 2.0419,
+      "step": 8215
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016255800007973902,
+      "loss": 2.09,
+      "step": 8216
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016267036375147722,
+      "loss": 4.7728,
+      "step": 8217
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016278259727179917,
+      "loss": 2.5372,
+      "step": 8218
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001628947004076225,
+      "loss": 1.0615,
+      "step": 8219
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016300667292613573,
+      "loss": 2.1976,
+      "step": 8220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016311851459479862,
+      "loss": 4.5971,
+      "step": 8221
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016323022518134264,
+      "loss": 2.935,
+      "step": 8222
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016334180445377153,
+      "loss": 2.1478,
+      "step": 8223
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016345325218036175,
+      "loss": 3.4121,
+      "step": 8224
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001635645681296629,
+      "loss": 4.296,
+      "step": 8225
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016367575207049825,
+      "loss": 5.0962,
+      "step": 8226
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016378680377196524,
+      "loss": 3.0024,
+      "step": 8227
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000163897723003436,
+      "loss": 1.1047,
+      "step": 8228
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001640085095345575,
+      "loss": 1.8277,
+      "step": 8229
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016411916313525282,
+      "loss": 1.9889,
+      "step": 8230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001642296835757206,
+      "loss": 2.9492,
+      "step": 8231
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016434007062643633,
+      "loss": 3.4291,
+      "step": 8232
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016445032405815231,
+      "loss": 3.8937,
+      "step": 8233
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016456044364189843,
+      "loss": 3.507,
+      "step": 8234
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016467042914898255,
+      "loss": 1.5539,
+      "step": 8235
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000164780280350991,
+      "loss": 3.0582,
+      "step": 8236
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016488999701978898,
+      "loss": 1.9353,
+      "step": 8237
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001649995789275211,
+      "loss": 1.8259,
+      "step": 8238
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001651090258466119,
+      "loss": 3.7529,
+      "step": 8239
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001652183375497661,
+      "loss": 1.8659,
+      "step": 8240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016532751380996942,
+      "loss": 2.3557,
+      "step": 8241
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016543655440048874,
+      "loss": 1.2544,
+      "step": 8242
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001655454590948727,
+      "loss": 2.8043,
+      "step": 8243
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016565422766695224,
+      "loss": 1.0035,
+      "step": 8244
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016576285989084095,
+      "loss": 3.2765,
+      "step": 8245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016587135554093556,
+      "loss": 3.9876,
+      "step": 8246
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016597971439191644,
+      "loss": 1.3232,
+      "step": 8247
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001660879362187481,
+      "loss": 1.2327,
+      "step": 8248
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016619602079667956,
+      "loss": 1.5802,
+      "step": 8249
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016630396790124488,
+      "loss": 2.041,
+      "step": 8250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016641177730826367,
+      "loss": 3.2636,
+      "step": 8251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016651944879384148,
+      "loss": 2.4962,
+      "step": 8252
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016662698213437026,
+      "loss": 2.8104,
+      "step": 8253
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001667343771065289,
+      "loss": 4.0972,
+      "step": 8254
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016684163348728364,
+      "loss": 4.6992,
+      "step": 8255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016694875105388848,
+      "loss": 4.5338,
+      "step": 8256
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016705572958388576,
+      "loss": 4.5508,
+      "step": 8257
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016716256885510657,
+      "loss": 4.1716,
+      "step": 8258
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016726926864567122,
+      "loss": 0.7115,
+      "step": 8259
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016737582873398963,
+      "loss": 3.8339,
+      "step": 8260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016748224889876178,
+      "loss": 2.3685,
+      "step": 8261
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016758852891897856,
+      "loss": 2.3068,
+      "step": 8262
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016769466857392156,
+      "loss": 4.1235,
+      "step": 8263
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016780066764316407,
+      "loss": 1.0732,
+      "step": 8264
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016790652590657117,
+      "loss": 2.3252,
+      "step": 8265
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016801224314430055,
+      "loss": 3.6548,
+      "step": 8266
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016811781913680265,
+      "loss": 3.0964,
+      "step": 8267
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016822325366482125,
+      "loss": 2.0534,
+      "step": 8268
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016832854650939406,
+      "loss": 2.8974,
+      "step": 8269
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016843369745185281,
+      "loss": 2.5956,
+      "step": 8270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001685387062738241,
+      "loss": 1.6869,
+      "step": 8271
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001686435727572296,
+      "loss": 1.4276,
+      "step": 8272
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001687482966842866,
+      "loss": 3.3781,
+      "step": 8273
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001688528778375085,
+      "loss": 1.6707,
+      "step": 8274
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001689573159997051,
+      "loss": 2.5688,
+      "step": 8275
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016906161095398326,
+      "loss": 3.7978,
+      "step": 8276
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016916576248374719,
+      "loss": 3.1716,
+      "step": 8277
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016926977037269892,
+      "loss": 0.8872,
+      "step": 8278
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016937363440483894,
+      "loss": 2.358,
+      "step": 8279
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001694773543644663,
+      "loss": 3.0242,
+      "step": 8280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016958093003617943,
+      "loss": 3.19,
+      "step": 8281
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016968436120487628,
+      "loss": 2.1859,
+      "step": 8282
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000169787647655755,
+      "loss": 3.5287,
+      "step": 8283
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00016989078917431418,
+      "loss": 1.5874,
+      "step": 8284
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001699937855463535,
+      "loss": 1.5467,
+      "step": 8285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000170096636557974,
+      "loss": 1.6915,
+      "step": 8286
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017019934199557867,
+      "loss": 2.4251,
+      "step": 8287
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017030190164587277,
+      "loss": 0.6365,
+      "step": 8288
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017040431529586432,
+      "loss": 2.7064,
+      "step": 8289
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017050658273286455,
+      "loss": 2.3767,
+      "step": 8290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017060870374448837,
+      "loss": 3.2889,
+      "step": 8291
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017071067811865468,
+      "loss": 1.4824,
+      "step": 8292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001708125056435872,
+      "loss": 3.7432,
+      "step": 8293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017091418610781428,
+      "loss": 2.0704,
+      "step": 8294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017101571930016988,
+      "loss": 2.0823,
+      "step": 8295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017111710500979373,
+      "loss": 1.1303,
+      "step": 8296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001712183430261318,
+      "loss": 3.4924,
+      "step": 8297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001713194331389369,
+      "loss": 3.0277,
+      "step": 8298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017142037513826898,
+      "loss": 2.9161,
+      "step": 8299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001715211688144955,
+      "loss": 4.0168,
+      "step": 8300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017162181395829196,
+      "loss": 5.63,
+      "step": 8301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001717223103606425,
+      "loss": 1.57,
+      "step": 8302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017182265781283995,
+      "loss": 1.8072,
+      "step": 8303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017192285610648656,
+      "loss": 4.5285,
+      "step": 8304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017202290503349433,
+      "loss": 2.5658,
+      "step": 8305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001721228043860855,
+      "loss": 2.7049,
+      "step": 8306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017222255395679293,
+      "loss": 5.1485,
+      "step": 8307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001723221535384605,
+      "loss": 2.9205,
+      "step": 8308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001724216029242436,
+      "loss": 4.2555,
+      "step": 8309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017252090190760953,
+      "loss": 1.9479,
+      "step": 8310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017262005028233798,
+      "loss": 1.5403,
+      "step": 8311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017271904784252145,
+      "loss": 0.9069,
+      "step": 8312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017281789438256548,
+      "loss": 2.3856,
+      "step": 8313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017291658969718942,
+      "loss": 3.9445,
+      "step": 8314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017301513358142657,
+      "loss": 4.0387,
+      "step": 8315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017311352583062483,
+      "loss": 2.4328,
+      "step": 8316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017321176624044687,
+      "loss": 3.3069,
+      "step": 8317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001733098546068708,
+      "loss": 3.9162,
+      "step": 8318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017340779072619043,
+      "loss": 2.1689,
+      "step": 8319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017350557439501578,
+      "loss": 4.1325,
+      "step": 8320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017360320541027346,
+      "loss": 3.9078,
+      "step": 8321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001737006835692071,
+      "loss": 3.3597,
+      "step": 8322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001737980086693778,
+      "loss": 2.1409,
+      "step": 8323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017389518050866436,
+      "loss": 2.2021,
+      "step": 8324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001739921988852643,
+      "loss": 2.2192,
+      "step": 8325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001740890635976934,
+      "loss": 1.9265,
+      "step": 8326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001741857744447868,
+      "loss": 2.091,
+      "step": 8327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017428233122569905,
+      "loss": 2.3122,
+      "step": 8328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001743787337399047,
+      "loss": 1.2545,
+      "step": 8329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017447498178719872,
+      "loss": 3.5971,
+      "step": 8330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017457107516769688,
+      "loss": 3.1437,
+      "step": 8331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017466701368183606,
+      "loss": 3.9348,
+      "step": 8332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017476279713037486,
+      "loss": 4.4392,
+      "step": 8333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017485842531439385,
+      "loss": 3.3172,
+      "step": 8334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001749538980352961,
+      "loss": 3.6201,
+      "step": 8335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017504921509480748,
+      "loss": 3.68,
+      "step": 8336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017514437629497718,
+      "loss": 2.1249,
+      "step": 8337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017523938143817805,
+      "loss": 2.8886,
+      "step": 8338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017533423032710705,
+      "loss": 1.9155,
+      "step": 8339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017542892276478563,
+      "loss": 4.2872,
+      "step": 8340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017552345855456012,
+      "loss": 1.4536,
+      "step": 8341
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017561783750010226,
+      "loss": 3.4522,
+      "step": 8342
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017571205940540945,
+      "loss": 2.3862,
+      "step": 8343
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001758061240748052,
+      "loss": 6.492,
+      "step": 8344
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017590003131293967,
+      "loss": 3.2782,
+      "step": 8345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001759937809247898,
+      "loss": 3.4399,
+      "step": 8346
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017608737271566005,
+      "loss": 3.7252,
+      "step": 8347
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017618080649118255,
+      "loss": 3.53,
+      "step": 8348
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017627408205731762,
+      "loss": 2.8214,
+      "step": 8349
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001763671992203541,
+      "loss": 3.5403,
+      "step": 8350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017646015778690984,
+      "loss": 0.9418,
+      "step": 8351
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017655295756393202,
+      "loss": 3.1287,
+      "step": 8352
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017664559835869765,
+      "loss": 3.9327,
+      "step": 8353
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017673807997881384,
+      "loss": 0.5822,
+      "step": 8354
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001768304022322183,
+      "loss": 2.1485,
+      "step": 8355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001769225649271796,
+      "loss": 4.1151,
+      "step": 8356
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017701456787229798,
+      "loss": 3.1954,
+      "step": 8357
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017710641087650514,
+      "loss": 3.1201,
+      "step": 8358
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017719809374906506,
+      "loss": 1.9391,
+      "step": 8359
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017728961629957422,
+      "loss": 3.7676,
+      "step": 8360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001773809783379621,
+      "loss": 1.947,
+      "step": 8361
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001774721796744915,
+      "loss": 4.0213,
+      "step": 8362
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000177563220119759,
+      "loss": 3.0871,
+      "step": 8363
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017765409948469527,
+      "loss": 1.2589,
+      "step": 8364
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001777448175805655,
+      "loss": 6.8003,
+      "step": 8365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017783537421896982,
+      "loss": 2.0406,
+      "step": 8366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017792576921184366,
+      "loss": 3.1464,
+      "step": 8367
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001780160023714582,
+      "loss": 3.6532,
+      "step": 8368
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001781060735104206,
+      "loss": 3.9516,
+      "step": 8369
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017819598244167456,
+      "loss": 2.2265,
+      "step": 8370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001782857289785007,
+      "loss": 4.2635,
+      "step": 8371
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017837531293451683,
+      "loss": 2.168,
+      "step": 8372
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001784647341236784,
+      "loss": 2.5066,
+      "step": 8373
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001785539923602789,
+      "loss": 2.8056,
+      "step": 8374
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017864308745895024,
+      "loss": 1.7595,
+      "step": 8375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017873201923466312,
+      "loss": 0.2069,
+      "step": 8376
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017882078750272742,
+      "loss": 4.0482,
+      "step": 8377
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017890939207879263,
+      "loss": 2.9533,
+      "step": 8378
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001789978327788481,
+      "loss": 4.0419,
+      "step": 8379
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017908610941922359,
+      "loss": 3.2198,
+      "step": 8380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017917422181658955,
+      "loss": 0.9704,
+      "step": 8381
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017926216978795753,
+      "loss": 5.2804,
+      "step": 8382
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017934995315068051,
+      "loss": 2.8369,
+      "step": 8383
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001794375717224534,
+      "loss": 1.3375,
+      "step": 8384
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017952502532131328,
+      "loss": 1.917,
+      "step": 8385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017961231376563985,
+      "loss": 2.3225,
+      "step": 8386
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017969943687415568,
+      "loss": 5.0938,
+      "step": 8387
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017978639446592708,
+      "loss": 4.5131,
+      "step": 8388
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017987318636036376,
+      "loss": 2.8153,
+      "step": 8389
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017995981237721957,
+      "loss": 2.9455,
+      "step": 8390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018004627233659292,
+      "loss": 2.1048,
+      "step": 8391
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001801325660589271,
+      "loss": 5.0018,
+      "step": 8392
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018021869336501062,
+      "loss": 2.4107,
+      "step": 8393
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001803046540759775,
+      "loss": 4.125,
+      "step": 8394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001803904480133079,
+      "loss": 1.1715,
+      "step": 8395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018047607499882822,
+      "loss": 2.6109,
+      "step": 8396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001805615348547116,
+      "loss": 3.1527,
+      "step": 8397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018064682740347832,
+      "loss": 3.0276,
+      "step": 8398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018073195246799602,
+      "loss": 1.5542,
+      "step": 8399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018081690987148025,
+      "loss": 0.9292,
+      "step": 8400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001809016994374947,
+      "loss": 1.8204,
+      "step": 8401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001809863209899517,
+      "loss": 3.9996,
+      "step": 8402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018107077435311237,
+      "loss": 3.6238,
+      "step": 8403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018115505935158724,
+      "loss": 3.0864,
+      "step": 8404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001812391758103364,
+      "loss": 1.7497,
+      "step": 8405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018132312355467005,
+      "loss": 4.5399,
+      "step": 8406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001814069024102487,
+      "loss": 2.2379,
+      "step": 8407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018149051220308363,
+      "loss": 3.3292,
+      "step": 8408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001815739527595372,
+      "loss": 3.9167,
+      "step": 8409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001816572239063233,
+      "loss": 3.7862,
+      "step": 8410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018174032547050754,
+      "loss": 1.7898,
+      "step": 8411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001818232572795078,
+      "loss": 2.1081,
+      "step": 8412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018190601916109444,
+      "loss": 4.3558,
+      "step": 8413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001819886109433908,
+      "loss": 2.5641,
+      "step": 8414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018207103245487338,
+      "loss": 2.4716,
+      "step": 8415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001821532835243723,
+      "loss": 4.2943,
+      "step": 8416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018223536398107178,
+      "loss": 3.3215,
+      "step": 8417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001823172736545102,
+      "loss": 2.0969,
+      "step": 8418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018239901237458058,
+      "loss": 4.3194,
+      "step": 8419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001824805799715313,
+      "loss": 3.4544,
+      "step": 8420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018256197627596575,
+      "loss": 4.6184,
+      "step": 8421
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018264320111884322,
+      "loss": 2.9431,
+      "step": 8422
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018272425433147908,
+      "loss": 2.1038,
+      "step": 8423
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001828051357455451,
+      "loss": 4.4108,
+      "step": 8424
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001828858451930698,
+      "loss": 1.4262,
+      "step": 8425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000182966382506439,
+      "loss": 2.8451,
+      "step": 8426
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018304674751839576,
+      "loss": 4.196,
+      "step": 8427
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018312694006204119,
+      "loss": 3.1524,
+      "step": 8428
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018320695997083443,
+      "loss": 0.819,
+      "step": 8429
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018328680707859317,
+      "loss": 2.5346,
+      "step": 8430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018336648121949401,
+      "loss": 4.3154,
+      "step": 8431
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001834459822280728,
+      "loss": 0.551,
+      "step": 8432
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018352530993922481,
+      "loss": 2.022,
+      "step": 8433
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018360446418820532,
+      "loss": 5.4342,
+      "step": 8434
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018368344481062976,
+      "loss": 3.283,
+      "step": 8435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018376225164247416,
+      "loss": 3.3526,
+      "step": 8436
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018384088452007576,
+      "loss": 4.4059,
+      "step": 8437
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018391934328013248,
+      "loss": 5.0399,
+      "step": 8438
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018399762775970436,
+      "loss": 2.3789,
+      "step": 8439
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001840757377962131,
+      "loss": 3.1058,
+      "step": 8440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018415367322744278,
+      "loss": 2.6064,
+      "step": 8441
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001842314338915401,
+      "loss": 5.2918,
+      "step": 8442
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018430901962701465,
+      "loss": 3.3428,
+      "step": 8443
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018438643027273937,
+      "loss": 2.1475,
+      "step": 8444
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018446366566795077,
+      "loss": 2.2367,
+      "step": 8445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018454072565224933,
+      "loss": 2.7259,
+      "step": 8446
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018461761006559983,
+      "loss": 3.0096,
+      "step": 8447
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018469431874833166,
+      "loss": 1.2281,
+      "step": 8448
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018477085154113918,
+      "loss": 3.1513,
+      "step": 8449
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018484720828508186,
+      "loss": 2.7409,
+      "step": 8450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018492338882158518,
+      "loss": 2.4102,
+      "step": 8451
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018499939299244017,
+      "loss": 3.8894,
+      "step": 8452
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018507522063980432,
+      "loss": 2.8279,
+      "step": 8453
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001851508716062016,
+      "loss": 3.1864,
+      "step": 8454
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018522634573452307,
+      "loss": 3.8796,
+      "step": 8455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018530164286802687,
+      "loss": 3.0212,
+      "step": 8456
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018537676285033881,
+      "loss": 2.5205,
+      "step": 8457
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018545170552545262,
+      "loss": 1.43,
+      "step": 8458
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018552647073773014,
+      "loss": 2.7779,
+      "step": 8459
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018560105833190192,
+      "loss": 3.3896,
+      "step": 8460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018567546815306726,
+      "loss": 6.8668,
+      "step": 8461
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001857497000466946,
+      "loss": 2.5982,
+      "step": 8462
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018582375385862215,
+      "loss": 3.7295,
+      "step": 8463
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018589762943505765,
+      "loss": 5.1486,
+      "step": 8464
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001859713266225792,
+      "loss": 2.2759,
+      "step": 8465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018604484526813523,
+      "loss": 2.9692,
+      "step": 8466
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018611818521904508,
+      "loss": 3.7248,
+      "step": 8467
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018619134632299903,
+      "loss": 4.3759,
+      "step": 8468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001862643284280592,
+      "loss": 4.22,
+      "step": 8469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018633713138265884,
+      "loss": 2.6979,
+      "step": 8470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001864097550356036,
+      "loss": 1.8796,
+      "step": 8471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018648219923607156,
+      "loss": 3.0032,
+      "step": 8472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001865544638336133,
+      "loss": 4.0209,
+      "step": 8473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018662654867815248,
+      "loss": 2.4623,
+      "step": 8474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018669845361998606,
+      "loss": 3.4262,
+      "step": 8475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018677017850978462,
+      "loss": 1.0818,
+      "step": 8476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001868417231985926,
+      "loss": 1.6423,
+      "step": 8477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018691308753782877,
+      "loss": 1.8906,
+      "step": 8478
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018698427137928638,
+      "loss": 4.1249,
+      "step": 8479
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018705527457513353,
+      "loss": 3.2272,
+      "step": 8480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018712609697791354,
+      "loss": 3.3563,
+      "step": 8481
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018719673844054503,
+      "loss": 3.6011,
+      "step": 8482
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018726719881632272,
+      "loss": 1.6801,
+      "step": 8483
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018733747795891706,
+      "loss": 3.5646,
+      "step": 8484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018740757572237502,
+      "loss": 1.2683,
+      "step": 8485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018747749196112027,
+      "loss": 2.8165,
+      "step": 8486
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018754722652995342,
+      "loss": 2.8741,
+      "step": 8487
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018761677928405242,
+      "loss": 1.8819,
+      "step": 8488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001876861500789727,
+      "loss": 2.748,
+      "step": 8489
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001877553387706477,
+      "loss": 6.4694,
+      "step": 8490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018782434521538895,
+      "loss": 1.7561,
+      "step": 8491
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001878931692698865,
+      "loss": 1.5753,
+      "step": 8492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018796181079120926,
+      "loss": 3.1859,
+      "step": 8493
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018803026963680507,
+      "loss": 4.4616,
+      "step": 8494
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018809854566450126,
+      "loss": 1.702,
+      "step": 8495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018816663873250474,
+      "loss": 3.4962,
+      "step": 8496
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018823454869940242,
+      "loss": 2.9447,
+      "step": 8497
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018830227542416153,
+      "loss": 3.7158,
+      "step": 8498
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018836981876612966,
+      "loss": 3.4782,
+      "step": 8499
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018843717858503563,
+      "loss": 1.6375,
+      "step": 8500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018850435474098903,
+      "loss": 3.6166,
+      "step": 8501
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001885713470944809,
+      "loss": 3.2688,
+      "step": 8502
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018863815550638414,
+      "loss": 4.7492,
+      "step": 8503
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001887047798379536,
+      "loss": 2.5712,
+      "step": 8504
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018877121995082642,
+      "loss": 4.228,
+      "step": 8505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018883747570702227,
+      "loss": 2.6592,
+      "step": 8506
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018890354696894375,
+      "loss": 3.0386,
+      "step": 8507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018896943359937663,
+      "loss": 1.0651,
+      "step": 8508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018903513546149,
+      "loss": 2.5748,
+      "step": 8509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001891006524188368,
+      "loss": 2.7953,
+      "step": 8510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018916598433535393,
+      "loss": 2.9787,
+      "step": 8511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001892311310753625,
+      "loss": 3.5642,
+      "step": 8512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001892960925035682,
+      "loss": 2.6636,
+      "step": 8513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001893608684850618,
+      "loss": 3.7064,
+      "step": 8514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018942545888531894,
+      "loss": 3.2715,
+      "step": 8515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001894898635702007,
+      "loss": 3.8535,
+      "step": 8516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001895540824059539,
+      "loss": 3.4576,
+      "step": 8517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018961811525921137,
+      "loss": 0.9591,
+      "step": 8518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018968196199699205,
+      "loss": 1.9155,
+      "step": 8519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018974562248670154,
+      "loss": 3.4019,
+      "step": 8520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018980909659613212,
+      "loss": 2.3138,
+      "step": 8521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018987238419346323,
+      "loss": 1.8943,
+      "step": 8522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018993548514726157,
+      "loss": 4.6972,
+      "step": 8523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018999839932648153,
+      "loss": 1.1758,
+      "step": 8524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019006112660046532,
+      "loss": 4.9841,
+      "step": 8525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019012366683894333,
+      "loss": 2.4156,
+      "step": 8526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019018601991203438,
+      "loss": 2.7829,
+      "step": 8527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019024818569024602,
+      "loss": 2.7023,
+      "step": 8528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001903101640444747,
+      "loss": 2.7328,
+      "step": 8529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019037195484600616,
+      "loss": 0.9034,
+      "step": 8530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019043355796651557,
+      "loss": 4.2298,
+      "step": 8531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019049497327806804,
+      "loss": 2.7363,
+      "step": 8532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019055620065311861,
+      "loss": 3.2672,
+      "step": 8533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019061723996451248,
+      "loss": 3.4803,
+      "step": 8534
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001906780910854856,
+      "loss": 3.4669,
+      "step": 8535
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019073875388966465,
+      "loss": 2.9502,
+      "step": 8536
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001907992282510675,
+      "loss": 3.8246,
+      "step": 8537
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019085951404410327,
+      "loss": 0.629,
+      "step": 8538
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019091961114357274,
+      "loss": 0.9638,
+      "step": 8539
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019097951942466854,
+      "loss": 3.1277,
+      "step": 8540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019103923876297545,
+      "loss": 2.7856,
+      "step": 8541
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019109876903447063,
+      "loss": 2.9672,
+      "step": 8542
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001911581101155239,
+      "loss": 2.7555,
+      "step": 8543
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019121726188289797,
+      "loss": 4.3638,
+      "step": 8544
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001912762242137486,
+      "loss": 5.0346,
+      "step": 8545
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019133499698562525,
+      "loss": 3.5695,
+      "step": 8546
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001913935800764708,
+      "loss": 1.8027,
+      "step": 8547
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019145197336462212,
+      "loss": 2.9381,
+      "step": 8548
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019151017672881026,
+      "loss": 1.7802,
+      "step": 8549
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001915681900481607,
+      "loss": 3.7436,
+      "step": 8550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001916260132021936,
+      "loss": 3.0498,
+      "step": 8551
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019168364607082406,
+      "loss": 2.3585,
+      "step": 8552
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001917410885343623,
+      "loss": 3.4277,
+      "step": 8553
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019179834047351402,
+      "loss": 4.1499,
+      "step": 8554
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001918554017693806,
+      "loss": 2.5966,
+      "step": 8555
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019191227230345932,
+      "loss": 3.208,
+      "step": 8556
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019196895195764362,
+      "loss": 3.2775,
+      "step": 8557
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019202544061422335,
+      "loss": 4.4325,
+      "step": 8558
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019208173815588503,
+      "loss": 3.4899,
+      "step": 8559
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019213784446571209,
+      "loss": 4.6147,
+      "step": 8560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019219375942718507,
+      "loss": 2.3752,
+      "step": 8561
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019224948292418187,
+      "loss": 2.8821,
+      "step": 8562
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019230501484097817,
+      "loss": 1.7665,
+      "step": 8563
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019236035506224738,
+      "loss": 2.052,
+      "step": 8564
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019241550347306107,
+      "loss": 3.1337,
+      "step": 8565
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019247045995888904,
+      "loss": 4.4885,
+      "step": 8566
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019252522440559977,
+      "loss": 2.96,
+      "step": 8567
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001925797966994606,
+      "loss": 3.9616,
+      "step": 8568
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019263417672713785,
+      "loss": 3.1515,
+      "step": 8569
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019268836437569715,
+      "loss": 5.3811,
+      "step": 8570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019274235953260367,
+      "loss": 3.2889,
+      "step": 8571
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019279616208572233,
+      "loss": 2.199,
+      "step": 8572
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019284977192331808,
+      "loss": 1.0463,
+      "step": 8573
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000192903188934056,
+      "loss": 3.7845,
+      "step": 8574
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001929564130070017,
+      "loss": 3.0523,
+      "step": 8575
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019300944403162146,
+      "loss": 4.2235,
+      "step": 8576
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019306228189778252,
+      "loss": 0.7196,
+      "step": 8577
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019311492649575315,
+      "loss": 0.5903,
+      "step": 8578
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019316737771620315,
+      "loss": 2.2091,
+      "step": 8579
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019321963545020375,
+      "loss": 1.7189,
+      "step": 8580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019327169958922811,
+      "loss": 4.4121,
+      "step": 8581
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001933235700251514,
+      "loss": 5.1263,
+      "step": 8582
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019337524665025114,
+      "loss": 1.6779,
+      "step": 8583
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001934267293572072,
+      "loss": 1.9389,
+      "step": 8584
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019347801803910235,
+      "loss": 3.2462,
+      "step": 8585
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019352911258942214,
+      "loss": 2.6347,
+      "step": 8586
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019358001290205543,
+      "loss": 3.0559,
+      "step": 8587
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019363071887129433,
+      "loss": 0.9952,
+      "step": 8588
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019368123039183464,
+      "loss": 2.6444,
+      "step": 8589
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000193731547358776,
+      "loss": 4.3477,
+      "step": 8590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019378166966762205,
+      "loss": 2.6408,
+      "step": 8591
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019383159721428065,
+      "loss": 4.7205,
+      "step": 8592
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019388132989506422,
+      "loss": 0.7117,
+      "step": 8593
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019393086760668976,
+      "loss": 3.7558,
+      "step": 8594
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019398021024627938,
+      "loss": 2.0582,
+      "step": 8595
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019402935771136004,
+      "loss": 4.3135,
+      "step": 8596
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001940783098998643,
+      "loss": 1.289,
+      "step": 8597
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019412706671012998,
+      "loss": 3.0659,
+      "step": 8598
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019417562804090082,
+      "loss": 2.1282,
+      "step": 8599
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019422399379132662,
+      "loss": 4.6513,
+      "step": 8600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019427216386096313,
+      "loss": 2.1006,
+      "step": 8601
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019432013814977262,
+      "loss": 1.8819,
+      "step": 8602
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001943679165581239,
+      "loss": 3.6609,
+      "step": 8603
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019441549898679264,
+      "loss": 2.6557,
+      "step": 8604
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019446288533696144,
+      "loss": 1.0039,
+      "step": 8605
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019451007551022018,
+      "loss": 3.6701,
+      "step": 8606
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019455706940856607,
+      "loss": 3.0439,
+      "step": 8607
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001946038669344039,
+      "loss": 3.1515,
+      "step": 8608
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019465046799054653,
+      "loss": 1.7979,
+      "step": 8609
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019469687248021457,
+      "loss": 3.0928,
+      "step": 8610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000194743080307037,
+      "loss": 1.7223,
+      "step": 8611
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019478909137505112,
+      "loss": 3.9533,
+      "step": 8612
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001948349055887029,
+      "loss": 2.2977,
+      "step": 8613
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019488052285284712,
+      "loss": 2.9871,
+      "step": 8614
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019492594307274764,
+      "loss": 2.315,
+      "step": 8615
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019497116615407748,
+      "loss": 2.8025,
+      "step": 8616
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019501619200291908,
+      "loss": 3.261,
+      "step": 8617
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001950610205257644,
+      "loss": 1.5633,
+      "step": 8618
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019510565162951534,
+      "loss": 2.6654,
+      "step": 8619
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019515008522148373,
+      "loss": 3.9132,
+      "step": 8620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001951943212093916,
+      "loss": 3.3427,
+      "step": 8621
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019523835950137128,
+      "loss": 3.7217,
+      "step": 8622
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019528220000596575,
+      "loss": 3.9655,
+      "step": 8623
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019532584263212878,
+      "loss": 3.5713,
+      "step": 8624
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019536928728922494,
+      "loss": 2.7929,
+      "step": 8625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019541253388703008,
+      "loss": 2.3259,
+      "step": 8626
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019545558233573132,
+      "loss": 1.4123,
+      "step": 8627
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001954984325459273,
+      "loss": 3.7263,
+      "step": 8628
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019554108442862837,
+      "loss": 3.2588,
+      "step": 8629
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019558353789525662,
+      "loss": 3.8933,
+      "step": 8630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001956257928576464,
+      "loss": 2.6042,
+      "step": 8631
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019566784922804418,
+      "loss": 3.7719,
+      "step": 8632
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001957097069191089,
+      "loss": 4.8866,
+      "step": 8633
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019575136584391216,
+      "loss": 2.8504,
+      "step": 8634
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001957928259159382,
+      "loss": 3.4454,
+      "step": 8635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019583408704908445,
+      "loss": 1.1139,
+      "step": 8636
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019587514915766124,
+      "loss": 1.5968,
+      "step": 8637
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019591601215639246,
+      "loss": 2.0369,
+      "step": 8638
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019595667596041528,
+      "loss": 2.8529,
+      "step": 8639
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019599714048528074,
+      "loss": 4.7907,
+      "step": 8640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001960374056469537,
+      "loss": 2.045,
+      "step": 8641
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019607747136181293,
+      "loss": 1.9276,
+      "step": 8642
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019611733754665157,
+      "loss": 2.4421,
+      "step": 8643
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019615700411867697,
+      "loss": 3.5651,
+      "step": 8644
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019619647099551114,
+      "loss": 3.8413,
+      "step": 8645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019623573809519077,
+      "loss": 4.3836,
+      "step": 8646
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019627480533616746,
+      "loss": 0.9056,
+      "step": 8647
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019631367263730787,
+      "loss": 2.7921,
+      "step": 8648
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019635233991789386,
+      "loss": 3.1079,
+      "step": 8649
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019639080709762268,
+      "loss": 3.5652,
+      "step": 8650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019642907409660719,
+      "loss": 1.653,
+      "step": 8651
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019646714083537594,
+      "loss": 4.4073,
+      "step": 8652
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019650500723487335,
+      "loss": 2.7299,
+      "step": 8653
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019654267321645997,
+      "loss": 3.2999,
+      "step": 8654
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019658013870191255,
+      "loss": 0.5045,
+      "step": 8655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019661740361342413,
+      "loss": 2.5477,
+      "step": 8656
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019665446787360438,
+      "loss": 1.6539,
+      "step": 8657
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019669133140547972,
+      "loss": 2.659,
+      "step": 8658
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019672799413249337,
+      "loss": 1.6863,
+      "step": 8659
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001967644559785055,
+      "loss": 4.8682,
+      "step": 8660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019680071686779366,
+      "loss": 3.9451,
+      "step": 8661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019683677672505253,
+      "loss": 4.019,
+      "step": 8662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019687263547539437,
+      "loss": 2.944,
+      "step": 8663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019690829304434916,
+      "loss": 2.9402,
+      "step": 8664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019694374935786457,
+      "loss": 4.2777,
+      "step": 8665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001969790043423063,
+      "loss": 3.3389,
+      "step": 8666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019701405792445817,
+      "loss": 2.8442,
+      "step": 8667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001970489100315222,
+      "loss": 3.6708,
+      "step": 8668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001970835605911189,
+      "loss": 3.0923,
+      "step": 8669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001971180095312873,
+      "loss": 1.9326,
+      "step": 8670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001971522567804851,
+      "loss": 1.5989,
+      "step": 8671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019718630226758904,
+      "loss": 3.9508,
+      "step": 8672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001972201459218947,
+      "loss": 2.6217,
+      "step": 8673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019725378767311685,
+      "loss": 1.6514,
+      "step": 8674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001972872274513896,
+      "loss": 1.1109,
+      "step": 8675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001973204651872665,
+      "loss": 3.3671,
+      "step": 8676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019735350081172067,
+      "loss": 3.1894,
+      "step": 8677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000197386334256145,
+      "loss": 2.7126,
+      "step": 8678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019741896545235227,
+      "loss": 1.9377,
+      "step": 8679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019745139433257524,
+      "loss": 3.6286,
+      "step": 8680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019748362082946686,
+      "loss": 4.5474,
+      "step": 8681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019751564487610038,
+      "loss": 3.2933,
+      "step": 8682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001975474664059694,
+      "loss": 2.8429,
+      "step": 8683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019757908535298832,
+      "loss": 2.4697,
+      "step": 8684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019761050165149205,
+      "loss": 3.9919,
+      "step": 8685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001976417152362364,
+      "loss": 3.4688,
+      "step": 8686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019767272604239821,
+      "loss": 3.2899,
+      "step": 8687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019770353400557544,
+      "loss": 2.4368,
+      "step": 8688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019773413906178723,
+      "loss": 2.1795,
+      "step": 8689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019776454114747426,
+      "loss": 2.646,
+      "step": 8690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001977947401994986,
+      "loss": 5.1374,
+      "step": 8691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000197824736155144,
+      "loss": 2.1968,
+      "step": 8692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019785452895211604,
+      "loss": 2.5501,
+      "step": 8693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019788411852854213,
+      "loss": 4.0208,
+      "step": 8694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019791350482297178,
+      "loss": 2.0727,
+      "step": 8695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019794268777437667,
+      "loss": 2.1422,
+      "step": 8696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001979716673221508,
+      "loss": 1.0814,
+      "step": 8697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019800044340611043,
+      "loss": 3.9625,
+      "step": 8698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019802901596649463,
+      "loss": 2.1648,
+      "step": 8699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001980573849439649,
+      "loss": 0.806,
+      "step": 8700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019808555027960568,
+      "loss": 3.6244,
+      "step": 8701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001981135119149243,
+      "loss": 0.4127,
+      "step": 8702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000198141269791851,
+      "loss": 2.7892,
+      "step": 8703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019816882385273942,
+      "loss": 2.0705,
+      "step": 8704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019819617404036625,
+      "loss": 4.0815,
+      "step": 8705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001982233202979317,
+      "loss": 0.7234,
+      "step": 8706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019825026256905945,
+      "loss": 2.4966,
+      "step": 8707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019827700079779682,
+      "loss": 1.4263,
+      "step": 8708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019830353492861487,
+      "loss": 2.9774,
+      "step": 8709
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019832986490640856,
+      "loss": 3.0192,
+      "step": 8710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019835599067649678,
+      "loss": 2.0836,
+      "step": 8711
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019838191218462248,
+      "loss": 3.4387,
+      "step": 8712
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019840762937695295,
+      "loss": 3.462,
+      "step": 8713
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019843314220007968,
+      "loss": 2.9434,
+      "step": 8714
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019845845060101856,
+      "loss": 4.7395,
+      "step": 8715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019848355452721012,
+      "loss": 2.1967,
+      "step": 8716
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001985084539265195,
+      "loss": 3.0028,
+      "step": 8717
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019853314874723657,
+      "loss": 2.1582,
+      "step": 8718
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000198557638938076,
+      "loss": 2.677,
+      "step": 8719
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019858192444817757,
+      "loss": 2.8301,
+      "step": 8720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019860600522710606,
+      "loss": 4.5855,
+      "step": 8721
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001986298812248514,
+      "loss": 3.24,
+      "step": 8722
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001986535523918289,
+      "loss": 4.5591,
+      "step": 8723
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019867701867887908,
+      "loss": 3.8367,
+      "step": 8724
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001987002800372681,
+      "loss": 3.1322,
+      "step": 8725
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001987233364186877,
+      "loss": 4.462,
+      "step": 8726
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019874618777525517,
+      "loss": 2.2202,
+      "step": 8727
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019876883405951377,
+      "loss": 2.1303,
+      "step": 8728
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019879127522443257,
+      "loss": 3.3684,
+      "step": 8729
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019881351122340654,
+      "loss": 1.252,
+      "step": 8730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019883554201025687,
+      "loss": 2.1193,
+      "step": 8731
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019885736753923086,
+      "loss": 3.0454,
+      "step": 8732
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019887898776500205,
+      "loss": 2.4638,
+      "step": 8733
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019890040264267035,
+      "loss": 4.2835,
+      "step": 8734
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019892161212776226,
+      "loss": 2.2364,
+      "step": 8735
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019894261617623072,
+      "loss": 1.9812,
+      "step": 8736
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019896341474445522,
+      "loss": 1.8174,
+      "step": 8737
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001989840077892422,
+      "loss": 2.0959,
+      "step": 8738
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019900439526782479,
+      "loss": 1.9588,
+      "step": 8739
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000199024577137863,
+      "loss": 1.445,
+      "step": 8740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019904455335744393,
+      "loss": 2.796,
+      "step": 8741
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019906432388508175,
+      "loss": 4.0052,
+      "step": 8742
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019908388867971772,
+      "loss": 3.9554,
+      "step": 8743
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019910324770072048,
+      "loss": 4.0565,
+      "step": 8744
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019912240090788595,
+      "loss": 3.1059,
+      "step": 8745
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019914134826143744,
+      "loss": 2.5948,
+      "step": 8746
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019916008972202585,
+      "loss": 4.2191,
+      "step": 8747
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019917862525072966,
+      "loss": 2.402,
+      "step": 8748
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001991969548090549,
+      "loss": 1.6844,
+      "step": 8749
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019921507835893553,
+      "loss": 2.7206,
+      "step": 8750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019923299586273324,
+      "loss": 4.9205,
+      "step": 8751
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019925070728323756,
+      "loss": 2.4516,
+      "step": 8752
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001992682125836662,
+      "loss": 3.0359,
+      "step": 8753
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019928551172766477,
+      "loss": 2.425,
+      "step": 8754
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019930260467930703,
+      "loss": 2.3018,
+      "step": 8755
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001993194914030951,
+      "loss": 1.6179,
+      "step": 8756
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001993361718639592,
+      "loss": 4.2112,
+      "step": 8757
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019935264602725797,
+      "loss": 1.8026,
+      "step": 8758
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001993689138587785,
+      "loss": 3.5633,
+      "step": 8759
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019938497532473646,
+      "loss": 2.1507,
+      "step": 8760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019940083039177593,
+      "loss": 3.1751,
+      "step": 8761
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019941647902696973,
+      "loss": 3.6659,
+      "step": 8762
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019943192119781938,
+      "loss": 4.7176,
+      "step": 8763
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001994471568722552,
+      "loss": 1.7458,
+      "step": 8764
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019946218601863625,
+      "loss": 1.0707,
+      "step": 8765
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019947700860575065,
+      "loss": 4.9695,
+      "step": 8766
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001994916246028154,
+      "loss": 2.4643,
+      "step": 8767
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001995060339794765,
+      "loss": 1.7873,
+      "step": 8768
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019952023670580916,
+      "loss": 3.8735,
+      "step": 8769
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001995342327523177,
+      "loss": 2.905,
+      "step": 8770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019954802208993564,
+      "loss": 3.3115,
+      "step": 8771
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019956160469002582,
+      "loss": 2.3147,
+      "step": 8772
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019957498052438046,
+      "loss": 1.9159,
+      "step": 8773
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019958814956522108,
+      "loss": 2.4252,
+      "step": 8774
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001996011117851987,
+      "loss": 2.4438,
+      "step": 8775
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019961386715739395,
+      "loss": 1.1746,
+      "step": 8776
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019962641565531692,
+      "loss": 1.6744,
+      "step": 8777
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019963875725290738,
+      "loss": 2.0518,
+      "step": 8778
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019965089192453475,
+      "loss": 1.8745,
+      "step": 8779
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019966281964499825,
+      "loss": 4.4952,
+      "step": 8780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019967454038952677,
+      "loss": 3.3652,
+      "step": 8781
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019968605413377925,
+      "loss": 2.8781,
+      "step": 8782
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001996973608538442,
+      "loss": 2.8164,
+      "step": 8783
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001997084605262404,
+      "loss": 4.2758,
+      "step": 8784
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019971935312791645,
+      "loss": 2.0476,
+      "step": 8785
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019973003863625098,
+      "loss": 2.6545,
+      "step": 8786
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019974051702905277,
+      "loss": 3.9035,
+      "step": 8787
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019975078828456062,
+      "loss": 2.5481,
+      "step": 8788
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001997608523814437,
+      "loss": 2.5342,
+      "step": 8789
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019977070929880116,
+      "loss": 3.3852,
+      "step": 8790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001997803590161626,
+      "loss": 2.1123,
+      "step": 8791
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019978980151348782,
+      "loss": 3.9345,
+      "step": 8792
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019979903677116702,
+      "loss": 3.8925,
+      "step": 8793
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019980806477002078,
+      "loss": 1.9986,
+      "step": 8794
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019981688549130006,
+      "loss": 3.5641,
+      "step": 8795
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019982549891668639,
+      "loss": 3.2886,
+      "step": 8796
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019983390502829166,
+      "loss": 4.4955,
+      "step": 8797
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001998421038086584,
+      "loss": 5.5773,
+      "step": 8798
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019985009524075978,
+      "loss": 2.5222,
+      "step": 8799
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019985787930799939,
+      "loss": 2.1703,
+      "step": 8800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001998654559942116,
+      "loss": 4.5576,
+      "step": 8801
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019987282528366148,
+      "loss": 2.6485,
+      "step": 8802
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019987998716104472,
+      "loss": 2.9464,
+      "step": 8803
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019988694161148787,
+      "loss": 3.1352,
+      "step": 8804
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019989368862054812,
+      "loss": 2.6251,
+      "step": 8805
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019990022817421358,
+      "loss": 1.3706,
+      "step": 8806
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019990656025890315,
+      "loss": 2.9569,
+      "step": 8807
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019991268486146662,
+      "loss": 3.8519,
+      "step": 8808
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999186019691846,
+      "loss": 1.9304,
+      "step": 8809
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999243115697687,
+      "loss": 1.6024,
+      "step": 8810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019992981365136143,
+      "loss": 3.6278,
+      "step": 8811
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999351082025363,
+      "loss": 3.7601,
+      "step": 8812
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019994019521229776,
+      "loss": 3.6847,
+      "step": 8813
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019994507467008133,
+      "loss": 1.2357,
+      "step": 8814
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999497465657535,
+      "loss": 3.8636,
+      "step": 8815
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999542108896119,
+      "loss": 2.4859,
+      "step": 8816
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019995846763238512,
+      "loss": 3.2125,
+      "step": 8817
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000199962516785233,
+      "loss": 1.6872,
+      "step": 8818
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019996635833974637,
+      "loss": 2.3931,
+      "step": 8819
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019996999228794727,
+      "loss": 2.644,
+      "step": 8820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019997341862228884,
+      "loss": 2.9865,
+      "step": 8821
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019997663733565537,
+      "loss": 4.1882,
+      "step": 8822
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999796484213624,
+      "loss": 3.3069,
+      "step": 8823
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019998245187315662,
+      "loss": 2.4054,
+      "step": 8824
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019998504768521587,
+      "loss": 1.9762,
+      "step": 8825
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019998743585214932,
+      "loss": 4.3693,
+      "step": 8826
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019998961636899734,
+      "loss": 1.599,
+      "step": 8827
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999158923123144,
+      "loss": 3.0084,
+      "step": 8828
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999335443475453,
+      "loss": 2.7662,
+      "step": 8829
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999491197590066,
+      "loss": 2.449,
+      "step": 8830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999626185143518,
+      "loss": 2.058,
+      "step": 8831
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999740405855478,
+      "loss": 3.6088,
+      "step": 8832
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999983385948873,
+      "loss": 4.1574,
+      "step": 8833
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999906545849197,
+      "loss": 1.5755,
+      "step": 8834
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999958464785923,
+      "loss": 2.4924,
+      "step": 8835
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999998961619109,
+      "loss": 1.6423,
+      "step": 8836
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002,
+      "loss": 3.9678,
+      "step": 8837
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999998961619109,
+      "loss": 2.6255,
+      "step": 8838
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999958464785923,
+      "loss": 3.8256,
+      "step": 8839
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999906545849197,
+      "loss": 3.3234,
+      "step": 8840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999983385948873,
+      "loss": 3.0024,
+      "step": 8841
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999740405855478,
+      "loss": 2.0196,
+      "step": 8842
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999626185143518,
+      "loss": 2.0582,
+      "step": 8843
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999491197590066,
+      "loss": 4.327,
+      "step": 8844
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999335443475453,
+      "loss": 0.8499,
+      "step": 8845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019999158923123144,
+      "loss": 2.8998,
+      "step": 8846
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019998961636899734,
+      "loss": 2.3588,
+      "step": 8847
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019998743585214932,
+      "loss": 3.7616,
+      "step": 8848
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019998504768521587,
+      "loss": 1.8877,
+      "step": 8849
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999824518731566,
+      "loss": 4.8314,
+      "step": 8850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999796484213624,
+      "loss": 4.0985,
+      "step": 8851
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019997663733565537,
+      "loss": 2.8349,
+      "step": 8852
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019997341862228884,
+      "loss": 2.4628,
+      "step": 8853
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019996999228794727,
+      "loss": 1.3244,
+      "step": 8854
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019996635833974637,
+      "loss": 1.4198,
+      "step": 8855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000199962516785233,
+      "loss": 3.1813,
+      "step": 8856
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019995846763238512,
+      "loss": 2.9796,
+      "step": 8857
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999542108896119,
+      "loss": 0.5755,
+      "step": 8858
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999497465657535,
+      "loss": 2.4975,
+      "step": 8859
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019994507467008133,
+      "loss": 3.8133,
+      "step": 8860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019994019521229776,
+      "loss": 2.8624,
+      "step": 8861
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999351082025363,
+      "loss": 2.2507,
+      "step": 8862
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019992981365136146,
+      "loss": 2.941,
+      "step": 8863
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001999243115697687,
+      "loss": 0.8507,
+      "step": 8864
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019991860196918462,
+      "loss": 3.1377,
+      "step": 8865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019991268486146662,
+      "loss": 4.3937,
+      "step": 8866
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019990656025890315,
+      "loss": 4.4194,
+      "step": 8867
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019990022817421358,
+      "loss": 3.5913,
+      "step": 8868
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019989368862054812,
+      "loss": 3.0267,
+      "step": 8869
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019988694161148787,
+      "loss": 5.0407,
+      "step": 8870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019987998716104472,
+      "loss": 1.6761,
+      "step": 8871
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019987282528366148,
+      "loss": 1.0662,
+      "step": 8872
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001998654559942116,
+      "loss": 3.4892,
+      "step": 8873
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019985787930799939,
+      "loss": 3.2294,
+      "step": 8874
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019985009524075978,
+      "loss": 3.9047,
+      "step": 8875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001998421038086584,
+      "loss": 0.6562,
+      "step": 8876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019983390502829166,
+      "loss": 1.3014,
+      "step": 8877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019982549891668639,
+      "loss": 2.8513,
+      "step": 8878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019981688549130006,
+      "loss": 5.0434,
+      "step": 8879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019980806477002078,
+      "loss": 2.9429,
+      "step": 8880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019979903677116702,
+      "loss": 2.3714,
+      "step": 8881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019978980151348782,
+      "loss": 3.6425,
+      "step": 8882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001997803590161626,
+      "loss": 5.0461,
+      "step": 8883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019977070929880116,
+      "loss": 3.1203,
+      "step": 8884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001997608523814437,
+      "loss": 1.2639,
+      "step": 8885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019975078828456068,
+      "loss": 3.7361,
+      "step": 8886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019974051702905277,
+      "loss": 2.9817,
+      "step": 8887
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000199730038636251,
+      "loss": 3.0787,
+      "step": 8888
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019971935312791645,
+      "loss": 3.5372,
+      "step": 8889
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019970846052624043,
+      "loss": 3.876,
+      "step": 8890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001996973608538442,
+      "loss": 2.4214,
+      "step": 8891
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019968605413377925,
+      "loss": 2.5616,
+      "step": 8892
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001996745403895268,
+      "loss": 1.812,
+      "step": 8893
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019966281964499825,
+      "loss": 1.9907,
+      "step": 8894
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019965089192453475,
+      "loss": 2.6831,
+      "step": 8895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019963875725290738,
+      "loss": 3.4247,
+      "step": 8896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019962641565531692,
+      "loss": 3.2259,
+      "step": 8897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019961386715739397,
+      "loss": 3.1009,
+      "step": 8898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001996011117851987,
+      "loss": 3.2218,
+      "step": 8899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019958814956522108,
+      "loss": 4.4034,
+      "step": 8900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019957498052438046,
+      "loss": 2.5911,
+      "step": 8901
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019956160469002582,
+      "loss": 3.2567,
+      "step": 8902
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019954802208993564,
+      "loss": 2.3073,
+      "step": 8903
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019953423275231768,
+      "loss": 1.5462,
+      "step": 8904
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019952023670580918,
+      "loss": 2.2672,
+      "step": 8905
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001995060339794765,
+      "loss": 0.943,
+      "step": 8906
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001994916246028154,
+      "loss": 4.3114,
+      "step": 8907
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019947700860575065,
+      "loss": 5.2063,
+      "step": 8908
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019946218601863628,
+      "loss": 3.8986,
+      "step": 8909
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019944715687225518,
+      "loss": 2.5453,
+      "step": 8910
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001994319211978194,
+      "loss": 2.9365,
+      "step": 8911
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019941647902696973,
+      "loss": 3.124,
+      "step": 8912
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019940083039177595,
+      "loss": 3.7008,
+      "step": 8913
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019938497532473646,
+      "loss": 4.7913,
+      "step": 8914
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001993689138587785,
+      "loss": 4.3362,
+      "step": 8915
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019935264602725795,
+      "loss": 2.0498,
+      "step": 8916
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001993361718639592,
+      "loss": 2.9901,
+      "step": 8917
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001993194914030951,
+      "loss": 3.0106,
+      "step": 8918
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019930260467930708,
+      "loss": 1.1191,
+      "step": 8919
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019928551172766477,
+      "loss": 5.7572,
+      "step": 8920
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019926821258366622,
+      "loss": 3.0687,
+      "step": 8921
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019925070728323756,
+      "loss": 1.3796,
+      "step": 8922
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019923299586273324,
+      "loss": 0.9914,
+      "step": 8923
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019921507835893556,
+      "loss": 1.8946,
+      "step": 8924
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019919695480905494,
+      "loss": 2.293,
+      "step": 8925
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001991786252507297,
+      "loss": 1.7509,
+      "step": 8926
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001991600897220259,
+      "loss": 3.9395,
+      "step": 8927
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019914134826143747,
+      "loss": 1.0349,
+      "step": 8928
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000199122400907886,
+      "loss": 3.8206,
+      "step": 8929
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001991032477007205,
+      "loss": 2.8274,
+      "step": 8930
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019908388867971778,
+      "loss": 0.6629,
+      "step": 8931
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019906432388508175,
+      "loss": 3.8358,
+      "step": 8932
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019904455335744396,
+      "loss": 2.0343,
+      "step": 8933
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000199024577137863,
+      "loss": 1.1617,
+      "step": 8934
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019900439526782479,
+      "loss": 5.9482,
+      "step": 8935
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019898400778924224,
+      "loss": 3.3469,
+      "step": 8936
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019896341474445525,
+      "loss": 4.1283,
+      "step": 8937
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019894261617623072,
+      "loss": 2.5151,
+      "step": 8938
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001989216121277623,
+      "loss": 4.0775,
+      "step": 8939
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019890040264267035,
+      "loss": 2.5304,
+      "step": 8940
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019887898776500202,
+      "loss": 1.837,
+      "step": 8941
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019885736753923086,
+      "loss": 4.4724,
+      "step": 8942
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019883554201025687,
+      "loss": 3.9405,
+      "step": 8943
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019881351122340654,
+      "loss": 0.3451,
+      "step": 8944
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019879127522443255,
+      "loss": 2.735,
+      "step": 8945
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001987688340595138,
+      "loss": 3.034,
+      "step": 8946
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019874618777525517,
+      "loss": 1.8512,
+      "step": 8947
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001987233364186877,
+      "loss": 1.5423,
+      "step": 8948
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001987002800372681,
+      "loss": 3.0294,
+      "step": 8949
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001986770186788791,
+      "loss": 2.5616,
+      "step": 8950
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001986535523918289,
+      "loss": 1.1729,
+      "step": 8951
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019862988122485144,
+      "loss": 5.6277,
+      "step": 8952
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019860600522710606,
+      "loss": 4.2364,
+      "step": 8953
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001985819244481776,
+      "loss": 2.467,
+      "step": 8954
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000198557638938076,
+      "loss": 2.948,
+      "step": 8955
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019853314874723657,
+      "loss": 1.8221,
+      "step": 8956
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019850845392651955,
+      "loss": 1.0038,
+      "step": 8957
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019848355452721015,
+      "loss": 2.3398,
+      "step": 8958
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019845845060101856,
+      "loss": 0.3946,
+      "step": 8959
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001984331422000797,
+      "loss": 2.673,
+      "step": 8960
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019840762937695295,
+      "loss": 0.8869,
+      "step": 8961
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019838191218462254,
+      "loss": 2.5445,
+      "step": 8962
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019835599067649678,
+      "loss": 1.463,
+      "step": 8963
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001983298649064086,
+      "loss": 4.3421,
+      "step": 8964
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001983035349286149,
+      "loss": 1.8404,
+      "step": 8965
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019827700079779682,
+      "loss": 4.9289,
+      "step": 8966
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019825026256905948,
+      "loss": 2.8095,
+      "step": 8967
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019822332029793168,
+      "loss": 3.2752,
+      "step": 8968
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019819617404036628,
+      "loss": 3.3686,
+      "step": 8969
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019816882385273945,
+      "loss": 1.1353,
+      "step": 8970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019814126979185104,
+      "loss": 4.2073,
+      "step": 8971
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001981135119149243,
+      "loss": 3.0781,
+      "step": 8972
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019808555027960573,
+      "loss": 0.2122,
+      "step": 8973
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001980573849439649,
+      "loss": 3.3427,
+      "step": 8974
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019802901596649463,
+      "loss": 2.0365,
+      "step": 8975
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001980004434061104,
+      "loss": 1.5409,
+      "step": 8976
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001979716673221508,
+      "loss": 4.5441,
+      "step": 8977
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019794268777437665,
+      "loss": 4.2972,
+      "step": 8978
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019791350482297178,
+      "loss": 3.5495,
+      "step": 8979
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019788411852854213,
+      "loss": 2.3606,
+      "step": 8980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019785452895211607,
+      "loss": 0.4499,
+      "step": 8981
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000197824736155144,
+      "loss": 3.7523,
+      "step": 8982
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019779474019949867,
+      "loss": 4.2336,
+      "step": 8983
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019776454114747426,
+      "loss": 1.4992,
+      "step": 8984
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001977341390617873,
+      "loss": 2.8224,
+      "step": 8985
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019770353400557546,
+      "loss": 2.7649,
+      "step": 8986
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019767272604239824,
+      "loss": 4.6402,
+      "step": 8987
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019764171523623642,
+      "loss": 2.2021,
+      "step": 8988
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019761050165149208,
+      "loss": 2.0958,
+      "step": 8989
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019757908535298838,
+      "loss": 1.4279,
+      "step": 8990
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001975474664059695,
+      "loss": 2.3109,
+      "step": 8991
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019751564487610038,
+      "loss": 3.6604,
+      "step": 8992
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019748362082946692,
+      "loss": 3.656,
+      "step": 8993
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019745139433257526,
+      "loss": 0.8747,
+      "step": 8994
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019741896545235235,
+      "loss": 2.2542,
+      "step": 8995
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019738633425614503,
+      "loss": 2.5772,
+      "step": 8996
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019735350081172073,
+      "loss": 4.7356,
+      "step": 8997
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001973204651872665,
+      "loss": 4.3883,
+      "step": 8998
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019728722745138957,
+      "loss": 2.2022,
+      "step": 8999
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019725378767311685,
+      "loss": 2.7753,
+      "step": 9000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001972201459218947,
+      "loss": 2.8117,
+      "step": 9001
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019718630226758907,
+      "loss": 4.8951,
+      "step": 9002
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019715225678048514,
+      "loss": 2.0162,
+      "step": 9003
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001971180095312873,
+      "loss": 3.1676,
+      "step": 9004
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001970835605911189,
+      "loss": 4.8401,
+      "step": 9005
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019704891003152224,
+      "loss": 2.3405,
+      "step": 9006
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019701405792445814,
+      "loss": 5.2586,
+      "step": 9007
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019697900434230634,
+      "loss": 2.0921,
+      "step": 9008
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019694374935786454,
+      "loss": 2.2117,
+      "step": 9009
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019690829304434918,
+      "loss": 4.0408,
+      "step": 9010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019687263547539437,
+      "loss": 3.3299,
+      "step": 9011
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019683677672505256,
+      "loss": 1.0517,
+      "step": 9012
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019680071686779366,
+      "loss": 3.3175,
+      "step": 9013
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019676445597850557,
+      "loss": 3.2619,
+      "step": 9014
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019672799413249337,
+      "loss": 0.9616,
+      "step": 9015
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001966913314054798,
+      "loss": 4.0074,
+      "step": 9016
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001966544678736044,
+      "loss": 1.6833,
+      "step": 9017
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019661740361342415,
+      "loss": 3.1366,
+      "step": 9018
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019658013870191258,
+      "loss": 4.8983,
+      "step": 9019
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019654267321646,
+      "loss": 1.1719,
+      "step": 9020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001965050072348734,
+      "loss": 3.5896,
+      "step": 9021
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000196467140835376,
+      "loss": 4.3479,
+      "step": 9022
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001964290740966072,
+      "loss": 3.8224,
+      "step": 9023
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019639080709762277,
+      "loss": 2.7989,
+      "step": 9024
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001963523399178939,
+      "loss": 1.8482,
+      "step": 9025
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019631367263730796,
+      "loss": 3.8517,
+      "step": 9026
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019627480533616752,
+      "loss": 0.452,
+      "step": 9027
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019623573809519088,
+      "loss": 1.9242,
+      "step": 9028
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019619647099551116,
+      "loss": 3.3406,
+      "step": 9029
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019615700411867694,
+      "loss": 4.3026,
+      "step": 9030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001961173375466516,
+      "loss": 2.9029,
+      "step": 9031
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019607747136181298,
+      "loss": 2.3157,
+      "step": 9032
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019603740564695372,
+      "loss": 2.8742,
+      "step": 9033
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019599714048528077,
+      "loss": 5.1302,
+      "step": 9034
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019595667596041528,
+      "loss": 5.0796,
+      "step": 9035
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001959160121563924,
+      "loss": 1.2381,
+      "step": 9036
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019587514915766127,
+      "loss": 3.4307,
+      "step": 9037
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019583408704908442,
+      "loss": 2.1171,
+      "step": 9038
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019579282591593823,
+      "loss": 2.8441,
+      "step": 9039
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001957513658439121,
+      "loss": 1.4131,
+      "step": 9040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019570970691910894,
+      "loss": 4.0431,
+      "step": 9041
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019566784922804415,
+      "loss": 3.8431,
+      "step": 9042
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019562579285764642,
+      "loss": 3.723,
+      "step": 9043
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001955835378952566,
+      "loss": 3.1623,
+      "step": 9044
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019554108442862837,
+      "loss": 3.7739,
+      "step": 9045
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001954984325459273,
+      "loss": 5.0123,
+      "step": 9046
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001954555823357314,
+      "loss": 0.4683,
+      "step": 9047
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019541253388703008,
+      "loss": 1.5367,
+      "step": 9048
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000195369287289225,
+      "loss": 3.0347,
+      "step": 9049
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019532584263212878,
+      "loss": 2.9634,
+      "step": 9050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001952822000059658,
+      "loss": 3.2075,
+      "step": 9051
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019523835950137128,
+      "loss": 2.7353,
+      "step": 9052
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019519432120939167,
+      "loss": 1.1222,
+      "step": 9053
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019515008522148376,
+      "loss": 3.4987,
+      "step": 9054
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019510565162951545,
+      "loss": 4.4936,
+      "step": 9055
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019506102052576445,
+      "loss": 1.9109,
+      "step": 9056
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019501619200291916,
+      "loss": 0.9607,
+      "step": 9057
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019497116615407753,
+      "loss": 2.2284,
+      "step": 9058
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019492594307274775,
+      "loss": 3.5672,
+      "step": 9059
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019488052285284715,
+      "loss": 1.859,
+      "step": 9060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019483490558870297,
+      "loss": 3.4266,
+      "step": 9061
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019478909137505112,
+      "loss": 2.3403,
+      "step": 9062
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000194743080307037,
+      "loss": 3.5592,
+      "step": 9063
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001946968724802146,
+      "loss": 2.9583,
+      "step": 9064
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019465046799054656,
+      "loss": 2.4378,
+      "step": 9065
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019460386693440394,
+      "loss": 3.5907,
+      "step": 9066
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000194557069408566,
+      "loss": 2.6641,
+      "step": 9067
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001945100755102202,
+      "loss": 3.4912,
+      "step": 9068
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001944628853369614,
+      "loss": 3.8385,
+      "step": 9069
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019441549898679264,
+      "loss": 2.8191,
+      "step": 9070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019436791655812386,
+      "loss": 3.1622,
+      "step": 9071
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019432013814977262,
+      "loss": 1.6276,
+      "step": 9072
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019427216386096307,
+      "loss": 3.9935,
+      "step": 9073
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019422399379132664,
+      "loss": 1.6994,
+      "step": 9074
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019417562804090082,
+      "loss": 2.9874,
+      "step": 9075
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019412706671013,
+      "loss": 4.9468,
+      "step": 9076
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001940783098998643,
+      "loss": 3.124,
+      "step": 9077
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019402935771136012,
+      "loss": 2.9974,
+      "step": 9078
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019398021024627938,
+      "loss": 3.1472,
+      "step": 9079
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019393086760668987,
+      "loss": 4.4115,
+      "step": 9080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019388132989506428,
+      "loss": 4.245,
+      "step": 9081
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001938315972142807,
+      "loss": 1.0744,
+      "step": 9082
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001937816696676221,
+      "loss": 3.3297,
+      "step": 9083
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019373154735877605,
+      "loss": 2.8607,
+      "step": 9084
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001936812303918347,
+      "loss": 2.0682,
+      "step": 9085
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019363071887129442,
+      "loss": 4.7077,
+      "step": 9086
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019358001290205545,
+      "loss": 2.753,
+      "step": 9087
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019352911258942225,
+      "loss": 3.4611,
+      "step": 9088
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019347801803910238,
+      "loss": 4.0555,
+      "step": 9089
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019342672935720732,
+      "loss": 1.8416,
+      "step": 9090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019337524665025116,
+      "loss": 2.3918,
+      "step": 9091
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001933235700251515,
+      "loss": 3.861,
+      "step": 9092
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019327169958922814,
+      "loss": 4.8793,
+      "step": 9093
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019321963545020372,
+      "loss": 2.8341,
+      "step": 9094
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019316737771620318,
+      "loss": 0.4311,
+      "step": 9095
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001931149264957532,
+      "loss": 4.84,
+      "step": 9096
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019306228189778252,
+      "loss": 3.47,
+      "step": 9097
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019300944403162148,
+      "loss": 5.2716,
+      "step": 9098
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019295641300700177,
+      "loss": 2.7536,
+      "step": 9099
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019290318893405597,
+      "loss": 3.6672,
+      "step": 9100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019284977192331814,
+      "loss": 2.9298,
+      "step": 9101
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001927961620857223,
+      "loss": 5.1757,
+      "step": 9102
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001927423595326037,
+      "loss": 2.4919,
+      "step": 9103
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019268836437569712,
+      "loss": 2.9781,
+      "step": 9104
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019263417672713787,
+      "loss": 3.1054,
+      "step": 9105
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019257979669946057,
+      "loss": 1.7174,
+      "step": 9106
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019252522440559982,
+      "loss": 3.2862,
+      "step": 9107
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000192470459958889,
+      "loss": 3.4361,
+      "step": 9108
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019241550347306113,
+      "loss": 4.5651,
+      "step": 9109
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019236035506224743,
+      "loss": 4.2539,
+      "step": 9110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019230501484097828,
+      "loss": 2.0076,
+      "step": 9111
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001922494829241819,
+      "loss": 2.4173,
+      "step": 9112
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019219375942718513,
+      "loss": 2.92,
+      "step": 9113
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001921378444657121,
+      "loss": 4.5726,
+      "step": 9114
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019208173815588506,
+      "loss": 3.1358,
+      "step": 9115
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019202544061422335,
+      "loss": 3.0913,
+      "step": 9116
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001919689519576437,
+      "loss": 4.7578,
+      "step": 9117
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019191227230345935,
+      "loss": 3.8661,
+      "step": 9118
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019185540176938073,
+      "loss": 3.5164,
+      "step": 9119
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019179834047351407,
+      "loss": 2.5937,
+      "step": 9120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019174108853436242,
+      "loss": 2.595,
+      "step": 9121
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001916836460708241,
+      "loss": 4.6024,
+      "step": 9122
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019162601320219373,
+      "loss": 1.5876,
+      "step": 9123
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019156819004816075,
+      "loss": 4.088,
+      "step": 9124
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019151017672881037,
+      "loss": 2.7709,
+      "step": 9125
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019145197336462215,
+      "loss": 2.0827,
+      "step": 9126
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019139358007647084,
+      "loss": 2.0288,
+      "step": 9127
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001913349969856253,
+      "loss": 3.2416,
+      "step": 9128
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019127622421374866,
+      "loss": 1.153,
+      "step": 9129
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019121726188289803,
+      "loss": 3.8852,
+      "step": 9130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019115811011552388,
+      "loss": 3.6275,
+      "step": 9131
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019109876903447069,
+      "loss": 1.7337,
+      "step": 9132
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019103923876297543,
+      "loss": 3.0888,
+      "step": 9133
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001909795194246686,
+      "loss": 3.5911,
+      "step": 9134
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001909196111435727,
+      "loss": 2.5205,
+      "step": 9135
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019085951404410333,
+      "loss": 3.593,
+      "step": 9136
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019079922825106747,
+      "loss": 2.9584,
+      "step": 9137
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001907387538896647,
+      "loss": 2.4241,
+      "step": 9138
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019067809108548554,
+      "loss": 2.1518,
+      "step": 9139
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019061723996451253,
+      "loss": 2.5016,
+      "step": 9140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001905562006531186,
+      "loss": 2.8599,
+      "step": 9141
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019049497327806815,
+      "loss": 0.963,
+      "step": 9142
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001904335579665156,
+      "loss": 3.834,
+      "step": 9143
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019037195484600621,
+      "loss": 2.6103,
+      "step": 9144
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019031016404447474,
+      "loss": 3.1731,
+      "step": 9145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019024818569024604,
+      "loss": 3.0735,
+      "step": 9146
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001901860199120344,
+      "loss": 2.7559,
+      "step": 9147
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019012366683894346,
+      "loss": 2.4935,
+      "step": 9148
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019006112660046535,
+      "loss": 3.022,
+      "step": 9149
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018999839932648164,
+      "loss": 2.2154,
+      "step": 9150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018993548514726163,
+      "loss": 3.709,
+      "step": 9151
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018987238419346337,
+      "loss": 2.4751,
+      "step": 9152
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018980909659613217,
+      "loss": 3.5931,
+      "step": 9153
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018974562248670168,
+      "loss": 2.542,
+      "step": 9154
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018968196199699208,
+      "loss": 4.9227,
+      "step": 9155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018961811525921148,
+      "loss": 4.2869,
+      "step": 9156
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018955408240595395,
+      "loss": 2.5904,
+      "step": 9157
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018948986357020074,
+      "loss": 4.6799,
+      "step": 9158
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018942545888531894,
+      "loss": 3.1583,
+      "step": 9159
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018936086848506185,
+      "loss": 1.4372,
+      "step": 9160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018929609250356827,
+      "loss": 4.4098,
+      "step": 9161
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018923113107536244,
+      "loss": 4.0486,
+      "step": 9162
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018916598433535399,
+      "loss": 2.8919,
+      "step": 9163
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001891006524188368,
+      "loss": 1.939,
+      "step": 9164
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018903513546149006,
+      "loss": 2.4342,
+      "step": 9165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018896943359937658,
+      "loss": 3.4139,
+      "step": 9166
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001889035469689438,
+      "loss": 2.6943,
+      "step": 9167
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018883747570702224,
+      "loss": 3.7064,
+      "step": 9168
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018877121995082645,
+      "loss": 3.9036,
+      "step": 9169
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018870477983795356,
+      "loss": 2.4693,
+      "step": 9170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001886381555063842,
+      "loss": 3.0045,
+      "step": 9171
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018857134709448088,
+      "loss": 2.4395,
+      "step": 9172
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001885043547409891,
+      "loss": 2.8383,
+      "step": 9173
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018843717858503566,
+      "loss": 2.0657,
+      "step": 9174
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001883698187661298,
+      "loss": 2.8948,
+      "step": 9175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001883022754241616,
+      "loss": 2.4147,
+      "step": 9176
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018823454869940247,
+      "loss": 1.6863,
+      "step": 9177
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018816663873250477,
+      "loss": 3.7992,
+      "step": 9178
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001880985456645014,
+      "loss": 3.0156,
+      "step": 9179
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018803026963680513,
+      "loss": 2.548,
+      "step": 9180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001879618107912094,
+      "loss": 3.5498,
+      "step": 9181
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018789316926988656,
+      "loss": 4.0685,
+      "step": 9182
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018782434521538908,
+      "loss": 1.9708,
+      "step": 9183
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018775533877064772,
+      "loss": 2.7634,
+      "step": 9184
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018768615007897283,
+      "loss": 3.1944,
+      "step": 9185
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018761677928405245,
+      "loss": 1.8371,
+      "step": 9186
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018754722652995355,
+      "loss": 2.4816,
+      "step": 9187
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018747749196112033,
+      "loss": 2.5766,
+      "step": 9188
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018740757572237516,
+      "loss": 4.2035,
+      "step": 9189
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018733747795891709,
+      "loss": 2.7322,
+      "step": 9190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018726719881632275,
+      "loss": 3.6076,
+      "step": 9191
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018719673844054508,
+      "loss": 2.0548,
+      "step": 9192
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018712609697791352,
+      "loss": 4.109,
+      "step": 9193
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001870552745751336,
+      "loss": 2.8128,
+      "step": 9194
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018698427137928632,
+      "loss": 2.8451,
+      "step": 9195
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018691308753782882,
+      "loss": 4.4374,
+      "step": 9196
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018684172319859257,
+      "loss": 3.7751,
+      "step": 9197
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018677017850978467,
+      "loss": 2.9856,
+      "step": 9198
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018669845361998604,
+      "loss": 2.0743,
+      "step": 9199
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001866265486781525,
+      "loss": 2.853,
+      "step": 9200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018655446383361324,
+      "loss": 1.4269,
+      "step": 9201
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001864821992360716,
+      "loss": 2.3053,
+      "step": 9202
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018640975503560355,
+      "loss": 3.2984,
+      "step": 9203
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018633713138265886,
+      "loss": 2.2474,
+      "step": 9204
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018626432842805914,
+      "loss": 2.2216,
+      "step": 9205
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018619134632299916,
+      "loss": 2.5574,
+      "step": 9206
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001861181852190451,
+      "loss": 2.3948,
+      "step": 9207
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018604484526813528,
+      "loss": 5.0712,
+      "step": 9208
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018597132662257924,
+      "loss": 3.044,
+      "step": 9209
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001858976294350577,
+      "loss": 4.594,
+      "step": 9210
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018582375385862217,
+      "loss": 2.5242,
+      "step": 9211
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018574970004669477,
+      "loss": 4.8636,
+      "step": 9212
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018567546815306728,
+      "loss": 3.7851,
+      "step": 9213
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018560105833190205,
+      "loss": 1.4978,
+      "step": 9214
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018552647073773022,
+      "loss": 3.3525,
+      "step": 9215
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018545170552545278,
+      "loss": 1.5578,
+      "step": 9216
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001853767628503389,
+      "loss": 4.2841,
+      "step": 9217
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018530164286802703,
+      "loss": 3.4765,
+      "step": 9218
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018522634573452313,
+      "loss": 3.2233,
+      "step": 9219
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018515087160620177,
+      "loss": 1.1162,
+      "step": 9220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018507522063980435,
+      "loss": 3.8857,
+      "step": 9221
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018499939299244022,
+      "loss": 3.4394,
+      "step": 9222
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018492338882158523,
+      "loss": 3.6287,
+      "step": 9223
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018484720828508192,
+      "loss": 2.5111,
+      "step": 9224
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018477085154113923,
+      "loss": 3.669,
+      "step": 9225
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018469431874833163,
+      "loss": 2.009,
+      "step": 9226
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001846176100655999,
+      "loss": 3.7228,
+      "step": 9227
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001845407256522493,
+      "loss": 1.9439,
+      "step": 9228
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001844636656679508,
+      "loss": 0.6528,
+      "step": 9229
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018438643027273935,
+      "loss": 1.09,
+      "step": 9230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001843090196270147,
+      "loss": 2.3561,
+      "step": 9231
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018423143389154005,
+      "loss": 2.4591,
+      "step": 9232
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018415367322744284,
+      "loss": 4.6666,
+      "step": 9233
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018407573779621306,
+      "loss": 2.8623,
+      "step": 9234
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018399762775970442,
+      "loss": 1.1504,
+      "step": 9235
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018391934328013245,
+      "loss": 0.3347,
+      "step": 9236
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001838408845200758,
+      "loss": 1.8033,
+      "step": 9237
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001837622516424742,
+      "loss": 2.2237,
+      "step": 9238
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001836834448106298,
+      "loss": 3.4213,
+      "step": 9239
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018360446418820537,
+      "loss": 3.934,
+      "step": 9240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018352530993922487,
+      "loss": 1.9054,
+      "step": 9241
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018344598222807285,
+      "loss": 0.7087,
+      "step": 9242
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018336648121949418,
+      "loss": 1.7588,
+      "step": 9243
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018328680707859323,
+      "loss": 4.5688,
+      "step": 9244
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018320695997083453,
+      "loss": 3.9591,
+      "step": 9245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018312694006204124,
+      "loss": 4.916,
+      "step": 9246
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001830467475183959,
+      "loss": 0.6248,
+      "step": 9247
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018296638250643902,
+      "loss": 4.4159,
+      "step": 9248
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018288584519307,
+      "loss": 2.5937,
+      "step": 9249
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018280513574554514,
+      "loss": 4.175,
+      "step": 9250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018272425433147924,
+      "loss": 2.8537,
+      "step": 9251
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001826432011188433,
+      "loss": 1.6514,
+      "step": 9252
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018256197627596591,
+      "loss": 2.1431,
+      "step": 9253
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018248057997153136,
+      "loss": 2.7078,
+      "step": 9254
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018239901237458064,
+      "loss": 4.5006,
+      "step": 9255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018231727365451025,
+      "loss": 2.9636,
+      "step": 9256
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018223536398107176,
+      "loss": 3.1282,
+      "step": 9257
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018215328352437238,
+      "loss": 1.2564,
+      "step": 9258
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018207103245487333,
+      "loss": 2.3072,
+      "step": 9259
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018198861094339085,
+      "loss": 2.5409,
+      "step": 9260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018190601916109439,
+      "loss": 2.5644,
+      "step": 9261
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018182325727950788,
+      "loss": 2.8399,
+      "step": 9262
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001817403254705075,
+      "loss": 1.422,
+      "step": 9263
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018165722390632336,
+      "loss": 2.9162,
+      "step": 9264
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018157395275953718,
+      "loss": 2.5504,
+      "step": 9265
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018149051220308368,
+      "loss": 1.7175,
+      "step": 9266
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018140690241024864,
+      "loss": 1.3193,
+      "step": 9267
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001813231235546701,
+      "loss": 1.1963,
+      "step": 9268
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018123917581033637,
+      "loss": 1.4873,
+      "step": 9269
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001811550593515873,
+      "loss": 2.1509,
+      "step": 9270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018107077435311242,
+      "loss": 2.2466,
+      "step": 9271
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018098632098995175,
+      "loss": 2.6296,
+      "step": 9272
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001809016994374948,
+      "loss": 3.5127,
+      "step": 9273
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018081690987148044,
+      "loss": 3.3429,
+      "step": 9274
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018073195246799608,
+      "loss": 3.1843,
+      "step": 9275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001806468274034785,
+      "loss": 1.0538,
+      "step": 9276
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018056153485471166,
+      "loss": 2.2381,
+      "step": 9277
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018047607499882838,
+      "loss": 4.5952,
+      "step": 9278
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018039044801330796,
+      "loss": 2.6992,
+      "step": 9279
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018030465407597768,
+      "loss": 2.3264,
+      "step": 9280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018021869336501064,
+      "loss": 2.5627,
+      "step": 9281
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018013256605892725,
+      "loss": 3.3435,
+      "step": 9282
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000180046272336593,
+      "loss": 2.9294,
+      "step": 9283
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017995981237721974,
+      "loss": 3.6316,
+      "step": 9284
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017987318636036382,
+      "loss": 2.2011,
+      "step": 9285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017978639446592716,
+      "loss": 0.7367,
+      "step": 9286
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017969943687415576,
+      "loss": 2.4372,
+      "step": 9287
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001796123137656398,
+      "loss": 2.9836,
+      "step": 9288
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017952502532131334,
+      "loss": 3.1588,
+      "step": 9289
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017943757172245338,
+      "loss": 3.687,
+      "step": 9290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001793499531506806,
+      "loss": 4.0488,
+      "step": 9291
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017926216978795748,
+      "loss": 4.9209,
+      "step": 9292
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017917422181658963,
+      "loss": 4.7542,
+      "step": 9293
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017908610941922356,
+      "loss": 2.5074,
+      "step": 9294
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017899783277884815,
+      "loss": 3.529,
+      "step": 9295
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017890939207879257,
+      "loss": 0.6733,
+      "step": 9296
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001788207875027275,
+      "loss": 5.5104,
+      "step": 9297
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017873201923466307,
+      "loss": 2.4505,
+      "step": 9298
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001786430874589503,
+      "loss": 3.0925,
+      "step": 9299
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017855399236027885,
+      "loss": 2.3222,
+      "step": 9300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001784647341236785,
+      "loss": 1.6055,
+      "step": 9301
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001783753129345169,
+      "loss": 4.2246,
+      "step": 9302
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001782857289785008,
+      "loss": 2.4978,
+      "step": 9303
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017819598244167464,
+      "loss": 3.5171,
+      "step": 9304
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017810607351042066,
+      "loss": 2.4282,
+      "step": 9305
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017801600237145823,
+      "loss": 4.1548,
+      "step": 9306
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017792576921184385,
+      "loss": 2.3024,
+      "step": 9307
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017783537421896988,
+      "loss": 5.04,
+      "step": 9308
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017774481758056566,
+      "loss": 1.8198,
+      "step": 9309
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017765409948469532,
+      "loss": 3.2181,
+      "step": 9310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001775632201197592,
+      "loss": 5.1883,
+      "step": 9311
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001774721796744916,
+      "loss": 5.9135,
+      "step": 9312
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017738097833796228,
+      "loss": 1.9348,
+      "step": 9313
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017728961629957427,
+      "loss": 1.725,
+      "step": 9314
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017719809374906525,
+      "loss": 0.7244,
+      "step": 9315
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017710641087650522,
+      "loss": 3.7354,
+      "step": 9316
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017701456787229815,
+      "loss": 3.1823,
+      "step": 9317
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017692256492717965,
+      "loss": 2.6577,
+      "step": 9318
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017683040223221826,
+      "loss": 4.6425,
+      "step": 9319
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001767380799788139,
+      "loss": 3.506,
+      "step": 9320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017664559835869762,
+      "loss": 2.8737,
+      "step": 9321
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001765529575639321,
+      "loss": 4.2507,
+      "step": 9322
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001764601577869098,
+      "loss": 2.2281,
+      "step": 9323
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017636719922035416,
+      "loss": 4.513,
+      "step": 9324
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017627408205731757,
+      "loss": 3.3691,
+      "step": 9325
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017618080649118263,
+      "loss": 0.6636,
+      "step": 9326
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017608737271566,
+      "loss": 2.9025,
+      "step": 9327
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001759937809247899,
+      "loss": 2.6093,
+      "step": 9328
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017590003131293962,
+      "loss": 1.2155,
+      "step": 9329
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017580612407480528,
+      "loss": 2.8848,
+      "step": 9330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001757120594054094,
+      "loss": 3.0308,
+      "step": 9331
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017561783750010235,
+      "loss": 2.8689,
+      "step": 9332
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001755234585545602,
+      "loss": 3.2251,
+      "step": 9333
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017542892276478566,
+      "loss": 3.6527,
+      "step": 9334
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001753342303271071,
+      "loss": 1.9983,
+      "step": 9335
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001752393814381781,
+      "loss": 2.0113,
+      "step": 9336
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017514437629497724,
+      "loss": 3.5192,
+      "step": 9337
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017504921509480764,
+      "loss": 5.0172,
+      "step": 9338
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017495389803529615,
+      "loss": 2.4071,
+      "step": 9339
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000174858425314394,
+      "loss": 4.619,
+      "step": 9340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001747627971303749,
+      "loss": 2.053,
+      "step": 9341
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017466701368183625,
+      "loss": 3.9562,
+      "step": 9342
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017457107516769696,
+      "loss": 0.9173,
+      "step": 9343
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001744749817871989,
+      "loss": 1.6551,
+      "step": 9344
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017437873373990476,
+      "loss": 1.6869,
+      "step": 9345
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001742823312256992,
+      "loss": 3.2527,
+      "step": 9346
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017418577444478688,
+      "loss": 1.3019,
+      "step": 9347
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001740890635976936,
+      "loss": 1.0623,
+      "step": 9348
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017399219888526436,
+      "loss": 1.879,
+      "step": 9349
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001738951805086644,
+      "loss": 2.8427,
+      "step": 9350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017379800866937784,
+      "loss": 3.3208,
+      "step": 9351
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017370068356920704,
+      "loss": 3.2713,
+      "step": 9352
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017360320541027352,
+      "loss": 0.9466,
+      "step": 9353
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017350557439501572,
+      "loss": 2.8488,
+      "step": 9354
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001734077907261905,
+      "loss": 2.8774,
+      "step": 9355
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017330985460687075,
+      "loss": 4.101,
+      "step": 9356
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017321176624044695,
+      "loss": 2.3438,
+      "step": 9357
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017311352583062478,
+      "loss": 2.9396,
+      "step": 9358
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017301513358142665,
+      "loss": 2.8063,
+      "step": 9359
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017291658969718936,
+      "loss": 1.4618,
+      "step": 9360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017281789438256554,
+      "loss": 1.8336,
+      "step": 9361
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017271904784252136,
+      "loss": 3.5121,
+      "step": 9362
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017262005028233806,
+      "loss": 0.7091,
+      "step": 9363
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017252090190760948,
+      "loss": 2.4428,
+      "step": 9364
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017242160292424365,
+      "loss": 3.7319,
+      "step": 9365
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017232215353846057,
+      "loss": 2.7192,
+      "step": 9366
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017222255395679299,
+      "loss": 1.8064,
+      "step": 9367
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001721228043860856,
+      "loss": 2.5916,
+      "step": 9368
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017202290503349452,
+      "loss": 0.3684,
+      "step": 9369
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001719228561064866,
+      "loss": 2.2603,
+      "step": 9370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017182265781284014,
+      "loss": 3.0066,
+      "step": 9371
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017172231036064256,
+      "loss": 2.784,
+      "step": 9372
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017162181395829218,
+      "loss": 1.9574,
+      "step": 9373
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017152116881449553,
+      "loss": 3.0294,
+      "step": 9374
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017142037513826917,
+      "loss": 2.2185,
+      "step": 9375
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017131943313893698,
+      "loss": 2.8078,
+      "step": 9376
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000171218343026132,
+      "loss": 4.2703,
+      "step": 9377
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017111710500979376,
+      "loss": 1.9041,
+      "step": 9378
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017101571930017007,
+      "loss": 2.0771,
+      "step": 9379
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017091418610781436,
+      "loss": 4.2491,
+      "step": 9380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017081250564358736,
+      "loss": 1.3117,
+      "step": 9381
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017071067811865473,
+      "loss": 3.2381,
+      "step": 9382
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017060870374448832,
+      "loss": 3.7591,
+      "step": 9383
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001705065827328646,
+      "loss": 4.5891,
+      "step": 9384
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017040431529586424,
+      "loss": 3.2013,
+      "step": 9385
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017030190164587283,
+      "loss": 2.3234,
+      "step": 9386
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017019934199557861,
+      "loss": 2.2828,
+      "step": 9387
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001700966365579741,
+      "loss": 2.6092,
+      "step": 9388
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016999378554635345,
+      "loss": 1.105,
+      "step": 9389
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016989078917431426,
+      "loss": 1.5164,
+      "step": 9390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016978764765575495,
+      "loss": 2.6365,
+      "step": 9391
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016968436120487636,
+      "loss": 2.6057,
+      "step": 9392
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016958093003617938,
+      "loss": 2.8959,
+      "step": 9393
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016947735436446636,
+      "loss": 3.8707,
+      "step": 9394
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016937363440483886,
+      "loss": 2.8497,
+      "step": 9395
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000169269770372699,
+      "loss": 3.8022,
+      "step": 9396
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001691657624837472,
+      "loss": 4.235,
+      "step": 9397
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001690616109539833,
+      "loss": 2.9378,
+      "step": 9398
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016895731599970518,
+      "loss": 0.8116,
+      "step": 9399
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001688528778375087,
+      "loss": 5.1673,
+      "step": 9400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016874829668428668,
+      "loss": 2.5609,
+      "step": 9401
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001686435727572298,
+      "loss": 2.3788,
+      "step": 9402
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016853870627382418,
+      "loss": 3.5988,
+      "step": 9403
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000168433697451853,
+      "loss": 3.0224,
+      "step": 9404
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016832854650939411,
+      "loss": 3.3938,
+      "step": 9405
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016822325366482147,
+      "loss": 2.0248,
+      "step": 9406
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001681178191368027,
+      "loss": 3.3156,
+      "step": 9407
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016801224314430074,
+      "loss": 0.4327,
+      "step": 9408
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016790652590657125,
+      "loss": 4.4936,
+      "step": 9409
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016780066764316426,
+      "loss": 3.4009,
+      "step": 9410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016769466857392167,
+      "loss": 3.3123,
+      "step": 9411
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016758852891897875,
+      "loss": 3.3266,
+      "step": 9412
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016748224889876186,
+      "loss": 2.6367,
+      "step": 9413
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001673758287339897,
+      "loss": 4.3097,
+      "step": 9414
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001672692686456713,
+      "loss": 4.0801,
+      "step": 9415
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016716256885510652,
+      "loss": 3.0359,
+      "step": 9416
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016705572958388584,
+      "loss": 4.7259,
+      "step": 9417
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016694875105388842,
+      "loss": 2.2612,
+      "step": 9418
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001668416334872837,
+      "loss": 2.8087,
+      "step": 9419
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016673437710652884,
+      "loss": 0.4188,
+      "step": 9420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016662698213437035,
+      "loss": 4.8752,
+      "step": 9421
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016651944879384143,
+      "loss": 1.3832,
+      "step": 9422
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016641177730826375,
+      "loss": 4.8766,
+      "step": 9423
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001663039679012448,
+      "loss": 2.8074,
+      "step": 9424
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016619602079667958,
+      "loss": 3.795,
+      "step": 9425
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016608793621874802,
+      "loss": 3.9287,
+      "step": 9426
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016597971439191652,
+      "loss": 2.1866,
+      "step": 9427
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016587135554093564,
+      "loss": 2.861,
+      "step": 9428
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016576285989084103,
+      "loss": 3.4373,
+      "step": 9429
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016565422766695234,
+      "loss": 1.0521,
+      "step": 9430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016554545909487278,
+      "loss": 2.3678,
+      "step": 9431
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001654365544004888,
+      "loss": 1.1415,
+      "step": 9432
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001653275138099696,
+      "loss": 3.1493,
+      "step": 9433
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016521833754976618,
+      "loss": 3.3273,
+      "step": 9434
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001651090258466121,
+      "loss": 1.7526,
+      "step": 9435
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001649995789275212,
+      "loss": 2.5911,
+      "step": 9436
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016488999701978917,
+      "loss": 2.7536,
+      "step": 9437
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016478028035099107,
+      "loss": 4.1066,
+      "step": 9438
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016467042914898277,
+      "loss": 0.9099,
+      "step": 9439
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016456044364189848,
+      "loss": 3.208,
+      "step": 9440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001644503240581525,
+      "loss": 2.3444,
+      "step": 9441
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001643400706264364,
+      "loss": 1.8798,
+      "step": 9442
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016422968357572085,
+      "loss": 4.4147,
+      "step": 9443
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001641191631352529,
+      "loss": 3.4653,
+      "step": 9444
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016400850953455771,
+      "loss": 3.9353,
+      "step": 9445
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016389772300343607,
+      "loss": 3.4094,
+      "step": 9446
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001637868037719652,
+      "loss": 4.3971,
+      "step": 9447
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016367575207049833,
+      "loss": 1.8583,
+      "step": 9448
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016356456812966283,
+      "loss": 2.4516,
+      "step": 9449
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016345325218036183,
+      "loss": 4.168,
+      "step": 9450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016334180445377148,
+      "loss": 1.3609,
+      "step": 9451
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016323022518134272,
+      "loss": 1.0558,
+      "step": 9452
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016311851459479854,
+      "loss": 5.4417,
+      "step": 9453
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016300667292613578,
+      "loss": 2.0511,
+      "step": 9454
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016289470040762243,
+      "loss": 3.2544,
+      "step": 9455
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016278259727179923,
+      "loss": 1.9309,
+      "step": 9456
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016267036375147717,
+      "loss": 1.3398,
+      "step": 9457
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016255800007973907,
+      "loss": 1.9295,
+      "step": 9458
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016244550648993712,
+      "loss": 0.4411,
+      "step": 9459
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001623328832156938,
+      "loss": 2.1977,
+      "step": 9460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016222013049090088,
+      "loss": 3.0089,
+      "step": 9461
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001621072485497189,
+      "loss": 2.1412,
+      "step": 9462
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001619942376265767,
+      "loss": 2.6989,
+      "step": 9463
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016188109795617125,
+      "loss": 1.2837,
+      "step": 9464
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016176782977346627,
+      "loss": 2.0369,
+      "step": 9465
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001616544333136933,
+      "loss": 2.8899,
+      "step": 9466
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001615409088123493,
+      "loss": 3.366,
+      "step": 9467
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001614272565051981,
+      "loss": 4.4189,
+      "step": 9468
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016131347662826792,
+      "loss": 2.5681,
+      "step": 9469
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016119956941785304,
+      "loss": 2.2701,
+      "step": 9470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016108553511051104,
+      "loss": 4.4999,
+      "step": 9471
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016097137394306457,
+      "loss": 1.9065,
+      "step": 9472
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001608570861525986,
+      "loss": 1.5619,
+      "step": 9473
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016074267197646225,
+      "loss": 3.0277,
+      "step": 9474
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001606281316522659,
+      "loss": 2.906,
+      "step": 9475
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001605134654178831,
+      "loss": 2.1181,
+      "step": 9476
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016039867351144786,
+      "loss": 1.6064,
+      "step": 9477
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001602837561713558,
+      "loss": 2.7869,
+      "step": 9478
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00016016871363626288,
+      "loss": 3.8859,
+      "step": 9479
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001600535461450848,
+      "loss": 3.621,
+      "step": 9480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015993825393699753,
+      "loss": 2.091,
+      "step": 9481
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015982283725143493,
+      "loss": 1.8698,
+      "step": 9482
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001597072963280905,
+      "loss": 2.6995,
+      "step": 9483
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015959163140691468,
+      "loss": 2.7799,
+      "step": 9484
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015947584272811647,
+      "loss": 2.1031,
+      "step": 9485
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015935993053216085,
+      "loss": 1.4092,
+      "step": 9486
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001592438950597704,
+      "loss": 1.6786,
+      "step": 9487
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015912773655192263,
+      "loss": 1.9513,
+      "step": 9488
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015901145524985157,
+      "loss": 3.4548,
+      "step": 9489
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015889505139504525,
+      "loss": 2.9577,
+      "step": 9490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015877852522924737,
+      "loss": 1.0749,
+      "step": 9491
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001586618769944545,
+      "loss": 1.8916,
+      "step": 9492
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015854510693291747,
+      "loss": 1.9452,
+      "step": 9493
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001584282152871398,
+      "loss": 3.5459,
+      "step": 9494
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001583112022998778,
+      "loss": 3.803,
+      "step": 9495
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001581940682141391,
+      "loss": 1.2906,
+      "step": 9496
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015807681327318385,
+      "loss": 3.9082,
+      "step": 9497
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001579594377205221,
+      "loss": 2.5224,
+      "step": 9498
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001578419417999155,
+      "loss": 1.4851,
+      "step": 9499
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015772432575537445,
+      "loss": 1.4799,
+      "step": 9500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001576065898311601,
+      "loss": 3.4055,
+      "step": 9501
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015748873427178127,
+      "loss": 3.6573,
+      "step": 9502
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001573707593219965,
+      "loss": 2.8145,
+      "step": 9503
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015725266522681106,
+      "loss": 1.3011,
+      "step": 9504
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015713445223147887,
+      "loss": 4.1036,
+      "step": 9505
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015701612058149954,
+      "loss": 3.8812,
+      "step": 9506
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015689767052262034,
+      "loss": 2.2579,
+      "step": 9507
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001567791023008332,
+      "loss": 3.5424,
+      "step": 9508
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001566604161623767,
+      "loss": 2.3756,
+      "step": 9509
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015654161235373318,
+      "loss": 4.0095,
+      "step": 9510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001564226911216299,
+      "loss": 0.5642,
+      "step": 9511
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015630365271303835,
+      "loss": 3.0952,
+      "step": 9512
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015618449737517235,
+      "loss": 2.5739,
+      "step": 9513
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015606522535548977,
+      "loss": 2.4568,
+      "step": 9514
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015594583690168955,
+      "loss": 2.9202,
+      "step": 9515
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015582633226171372,
+      "loss": 2.654,
+      "step": 9516
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001557067116837443,
+      "loss": 1.222,
+      "step": 9517
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015558697541620534,
+      "loss": 0.9929,
+      "step": 9518
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015546712370775997,
+      "loss": 3.9093,
+      "step": 9519
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001553471568073122,
+      "loss": 1.725,
+      "step": 9520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015522707496400418,
+      "loss": 0.9601,
+      "step": 9521
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015510687842721778,
+      "loss": 0.9965,
+      "step": 9522
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015498656744657223,
+      "loss": 2.3978,
+      "step": 9523
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015486614227192485,
+      "loss": 2.8218,
+      "step": 9524
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001547456031533701,
+      "loss": 1.2054,
+      "step": 9525
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001546249503412391,
+      "loss": 2.95,
+      "step": 9526
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001545041840860986,
+      "loss": 2.8791,
+      "step": 9527
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015438330463875196,
+      "loss": 4.0493,
+      "step": 9528
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015426231225023634,
+      "loss": 1.0179,
+      "step": 9529
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001541412071718247,
+      "loss": 1.7885,
+      "step": 9530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015401998965502292,
+      "loss": 2.1518,
+      "step": 9531
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015389865995157141,
+      "loss": 0.4896,
+      "step": 9532
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015377721831344253,
+      "loss": 2.9593,
+      "step": 9533
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015365566499284222,
+      "loss": 2.4586,
+      "step": 9534
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015353400024220717,
+      "loss": 1.1056,
+      "step": 9535
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015341222431420668,
+      "loss": 2.5957,
+      "step": 9536
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015329033746173972,
+      "loss": 0.7196,
+      "step": 9537
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001531683399379369,
+      "loss": 1.607,
+      "step": 9538
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015304623199615734,
+      "loss": 1.7545,
+      "step": 9539
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015292401388999078,
+      "loss": 0.8801,
+      "step": 9540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001528016858732547,
+      "loss": 3.1614,
+      "step": 9541
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015267924819999538,
+      "loss": 0.5793,
+      "step": 9542
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015255670112448666,
+      "loss": 1.8644,
+      "step": 9543
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015243404490122953,
+      "loss": 1.9725,
+      "step": 9544
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015231127978495215,
+      "loss": 2.9417,
+      "step": 9545
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015218840603060786,
+      "loss": 0.7136,
+      "step": 9546
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015206542389337675,
+      "loss": 1.1432,
+      "step": 9547
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015194233362866283,
+      "loss": 1.0919,
+      "step": 9548
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015181913549209587,
+      "loss": 0.3247,
+      "step": 9549
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015169582973952842,
+      "loss": 2.4542,
+      "step": 9550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001515724166270378,
+      "loss": 2.0367,
+      "step": 9551
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015144889641092305,
+      "loss": 2.2845,
+      "step": 9552
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015132526934770674,
+      "loss": 2.6804,
+      "step": 9553
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015120153569413248,
+      "loss": 3.0552,
+      "step": 9554
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001510776957071657,
+      "loss": 1.2419,
+      "step": 9555
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015095374964399255,
+      "loss": 3.898,
+      "step": 9556
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001508296977620195,
+      "loss": 2.5482,
+      "step": 9557
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015070554031887273,
+      "loss": 2.531,
+      "step": 9558
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015058127757239787,
+      "loss": 2.7218,
+      "step": 9559
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001504569097806585,
+      "loss": 1.8681,
+      "step": 9560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015033243720193764,
+      "loss": 2.6994,
+      "step": 9561
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015020786009473444,
+      "loss": 1.2507,
+      "step": 9562
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00015008317871776657,
+      "loss": 4.2356,
+      "step": 9563
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014995839332996688,
+      "loss": 2.389,
+      "step": 9564
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014983350419048557,
+      "loss": 3.7089,
+      "step": 9565
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014970851155868698,
+      "loss": 1.2031,
+      "step": 9566
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001495834156941516,
+      "loss": 3.1334,
+      "step": 9567
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014945821685667323,
+      "loss": 1.964,
+      "step": 9568
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014933291530626057,
+      "loss": 3.3137,
+      "step": 9569
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001492075113031345,
+      "loss": 2.9622,
+      "step": 9570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014908200510772987,
+      "loss": 4.0155,
+      "step": 9571
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001489563969806927,
+      "loss": 2.7565,
+      "step": 9572
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001488306871828813,
+      "loss": 2.1001,
+      "step": 9573
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014870487597536496,
+      "loss": 4.1262,
+      "step": 9574
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014857896361942342,
+      "loss": 2.3456,
+      "step": 9575
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014845295037654716,
+      "loss": 2.4267,
+      "step": 9576
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000148326836508435,
+      "loss": 2.8124,
+      "step": 9577
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014820062227699604,
+      "loss": 1.8237,
+      "step": 9578
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014807430794434658,
+      "loss": 2.6111,
+      "step": 9579
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000147947893772812,
+      "loss": 1.2459,
+      "step": 9580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014782138002492377,
+      "loss": 5.5931,
+      "step": 9581
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014769476696342143,
+      "loss": 2.9379,
+      "step": 9582
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001475680548512496,
+      "loss": 4.0844,
+      "step": 9583
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014744124395155966,
+      "loss": 1.0233,
+      "step": 9584
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001473143345277071,
+      "loss": 2.6783,
+      "step": 9585
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014718732684325319,
+      "loss": 2.605,
+      "step": 9586
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001470602211619621,
+      "loss": 4.211,
+      "step": 9587
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014693301774780225,
+      "loss": 3.164,
+      "step": 9588
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014680571686494483,
+      "loss": 3.3662,
+      "step": 9589
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014667831877776362,
+      "loss": 2.9885,
+      "step": 9590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001465508237508336,
+      "loss": 3.5966,
+      "step": 9591
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014642323204893217,
+      "loss": 3.3132,
+      "step": 9592
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014629554393703634,
+      "loss": 3.2982,
+      "step": 9593
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001461677596803245,
+      "loss": 2.5333,
+      "step": 9594
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014603987954417348,
+      "loss": 3.2482,
+      "step": 9595
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014591190379416048,
+      "loss": 3.5621,
+      "step": 9596
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014578383269606002,
+      "loss": 1.861,
+      "step": 9597
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001456556665158459,
+      "loss": 2.9298,
+      "step": 9598
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014552740551968812,
+      "loss": 2.3816,
+      "step": 9599
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014539904997395482,
+      "loss": 1.5348,
+      "step": 9600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014527060014520927,
+      "loss": 1.6676,
+      "step": 9601
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014514205630021182,
+      "loss": 2.0485,
+      "step": 9602
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001450134187059169,
+      "loss": 4.1216,
+      "step": 9603
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014488468762947437,
+      "loss": 1.0565,
+      "step": 9604
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014475586333822794,
+      "loss": 5.4865,
+      "step": 9605
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014462694609971503,
+      "loss": 3.6313,
+      "step": 9606
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000144497936181666,
+      "loss": 1.0993,
+      "step": 9607
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014436883385200358,
+      "loss": 3.7083,
+      "step": 9608
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014423963937884303,
+      "loss": 2.0319,
+      "step": 9609
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001441103530304898,
+      "loss": 2.6691,
+      "step": 9610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014398097507544155,
+      "loss": 1.9133,
+      "step": 9611
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014385150578238482,
+      "loss": 4.2699,
+      "step": 9612
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001437219454201971,
+      "loss": 3.8417,
+      "step": 9613
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014359229425794378,
+      "loss": 2.6472,
+      "step": 9614
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014346255256488007,
+      "loss": 2.5681,
+      "step": 9615
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014333272061044792,
+      "loss": 0.7727,
+      "step": 9616
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000143202798664278,
+      "loss": 1.3391,
+      "step": 9617
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001430727869961868,
+      "loss": 5.8144,
+      "step": 9618
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014294268587617773,
+      "loss": 1.2106,
+      "step": 9619
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014281249557443978,
+      "loss": 4.466,
+      "step": 9620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001426822163613474,
+      "loss": 2.4539,
+      "step": 9621
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000142551848507459,
+      "loss": 3.7487,
+      "step": 9622
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014242139228351817,
+      "loss": 1.0067,
+      "step": 9623
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001422908479604508,
+      "loss": 4.668,
+      "step": 9624
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014216021580936697,
+      "loss": 3.6876,
+      "step": 9625
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001420294961015579,
+      "loss": 3.4192,
+      "step": 9626
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014189868910849791,
+      "loss": 3.2932,
+      "step": 9627
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014176779510184135,
+      "loss": 2.0164,
+      "step": 9628
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014163681435342454,
+      "loss": 1.6922,
+      "step": 9629
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014150574713526261,
+      "loss": 2.342,
+      "step": 9630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014137459371955165,
+      "loss": 2.3814,
+      "step": 9631
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014124335437866537,
+      "loss": 4.6197,
+      "step": 9632
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014111202938515724,
+      "loss": 2.7371,
+      "step": 9633
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001409806190117574,
+      "loss": 0.7108,
+      "step": 9634
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014084912353137452,
+      "loss": 1.889,
+      "step": 9635
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001407175432170929,
+      "loss": 3.445,
+      "step": 9636
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00014058587834217366,
+      "loss": 3.9392,
+      "step": 9637
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001404541291800534,
+      "loss": 3.6274,
+      "step": 9638
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00014032229600434356,
+      "loss": 2.2294,
+      "step": 9639
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00014019037908883075,
+      "loss": 2.2211,
+      "step": 9640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00014005837870747432,
+      "loss": 2.7268,
+      "step": 9641
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013992629513440825,
+      "loss": 3.1714,
+      "step": 9642
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013979412864393803,
+      "loss": 2.3569,
+      "step": 9643
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001396618795105426,
+      "loss": 4.3419,
+      "step": 9644
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013952954800887127,
+      "loss": 0.7534,
+      "step": 9645
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013939713441374574,
+      "loss": 0.1352,
+      "step": 9646
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013926463900015677,
+      "loss": 2.5124,
+      "step": 9647
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013913206204326648,
+      "loss": 3.8645,
+      "step": 9648
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013899940381840512,
+      "loss": 1.6128,
+      "step": 9649
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013886666460107239,
+      "loss": 4.5695,
+      "step": 9650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013873384466693602,
+      "loss": 1.0953,
+      "step": 9651
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013860094429183156,
+      "loss": 3.7365,
+      "step": 9652
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013846796375176088,
+      "loss": 3.498,
+      "step": 9653
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013833490332289359,
+      "loss": 1.9022,
+      "step": 9654
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013820176328156384,
+      "loss": 1.677,
+      "step": 9655
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001380685439042724,
+      "loss": 0.543,
+      "step": 9656
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013793524546768356,
+      "loss": 3.2658,
+      "step": 9657
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013780186824862708,
+      "loss": 1.9152,
+      "step": 9658
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000137668412524095,
+      "loss": 1.1194,
+      "step": 9659
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013753487857124375,
+      "loss": 2.2498,
+      "step": 9660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013740126666739083,
+      "loss": 5.5024,
+      "step": 9661
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013726757709001702,
+      "loss": 1.7558,
+      "step": 9662
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001371338101167631,
+      "loss": 2.6746,
+      "step": 9663
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013699996602543183,
+      "loss": 3.0797,
+      "step": 9664
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013686604509398483,
+      "loss": 3.7077,
+      "step": 9665
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013673204760054469,
+      "loss": 2.0683,
+      "step": 9666
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001365979738233917,
+      "loss": 3.2668,
+      "step": 9667
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013646382404096543,
+      "loss": 2.6116,
+      "step": 9668
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013632959853186294,
+      "loss": 2.4344,
+      "step": 9669
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001361952975748388,
+      "loss": 3.873,
+      "step": 9670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013606092144880367,
+      "loss": 1.9329,
+      "step": 9671
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013592647043282502,
+      "loss": 3.2004,
+      "step": 9672
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013579194480612596,
+      "loss": 1.3149,
+      "step": 9673
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001356573448480835,
+      "loss": 1.9852,
+      "step": 9674
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013552267083823032,
+      "loss": 2.9671,
+      "step": 9675
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013538792305625163,
+      "loss": 1.2124,
+      "step": 9676
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013525310178198714,
+      "loss": 2.6924,
+      "step": 9677
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001351182072954278,
+      "loss": 1.7241,
+      "step": 9678
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013498323987671805,
+      "loss": 1.246,
+      "step": 9679
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013484819980615255,
+      "loss": 3.0742,
+      "step": 9680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001347130873641775,
+      "loss": 3.0176,
+      "step": 9681
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001345779028313893,
+      "loss": 0.9356,
+      "step": 9682
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013444264648853398,
+      "loss": 3.7447,
+      "step": 9683
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013430731861650675,
+      "loss": 2.112,
+      "step": 9684
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013417191949635155,
+      "loss": 1.8217,
+      "step": 9685
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001340364494092596,
+      "loss": 3.35,
+      "step": 9686
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013390090863657064,
+      "loss": 3.0388,
+      "step": 9687
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013376529745976991,
+      "loss": 3.2816,
+      "step": 9688
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001336296161604902,
+      "loss": 1.0866,
+      "step": 9689
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013349386502050856,
+      "loss": 2.7202,
+      "step": 9690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013335804432174845,
+      "loss": 2.4638,
+      "step": 9691
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013322215434627643,
+      "loss": 2.6263,
+      "step": 9692
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013308619537630428,
+      "loss": 3.9505,
+      "step": 9693
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013295016769418575,
+      "loss": 1.6627,
+      "step": 9694
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013281407158241855,
+      "loss": 2.656,
+      "step": 9695
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013267790732364123,
+      "loss": 2.5217,
+      "step": 9696
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001325416752006352,
+      "loss": 1.4319,
+      "step": 9697
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013240537549632163,
+      "loss": 1.0918,
+      "step": 9698
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001322690084937627,
+      "loss": 1.3953,
+      "step": 9699
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013213257447616022,
+      "loss": 1.6203,
+      "step": 9700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013199607372685505,
+      "loss": 1.6459,
+      "step": 9701
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013185950652932682,
+      "loss": 4.5987,
+      "step": 9702
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013172287316719284,
+      "loss": 1.8329,
+      "step": 9703
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001315861739242086,
+      "loss": 3.817,
+      "step": 9704
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001314494090842652,
+      "loss": 0.7285,
+      "step": 9705
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013131257893139126,
+      "loss": 0.8021,
+      "step": 9706
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013117568374974975,
+      "loss": 0.5685,
+      "step": 9707
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013103872382364006,
+      "loss": 2.1492,
+      "step": 9708
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013090169943749463,
+      "loss": 2.9654,
+      "step": 9709
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013076461087588114,
+      "loss": 4.4815,
+      "step": 9710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013062745842349922,
+      "loss": 2.468,
+      "step": 9711
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013049024236518245,
+      "loss": 4.1145,
+      "step": 9712
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001303529629858955,
+      "loss": 2.2889,
+      "step": 9713
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013021562057073503,
+      "loss": 4.9486,
+      "step": 9714
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013007821540492854,
+      "loss": 1.3733,
+      "step": 9715
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012994074777383398,
+      "loss": 2.3966,
+      "step": 9716
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001298032179629384,
+      "loss": 2.681,
+      "step": 9717
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012966562625785907,
+      "loss": 2.7217,
+      "step": 9718
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001295279729443406,
+      "loss": 0.3693,
+      "step": 9719
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012939025830825673,
+      "loss": 2.4272,
+      "step": 9720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001292524826356073,
+      "loss": 1.2051,
+      "step": 9721
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012911464621252026,
+      "loss": 2.3132,
+      "step": 9722
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001289767493252483,
+      "loss": 3.4526,
+      "step": 9723
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012883879226017114,
+      "loss": 1.2038,
+      "step": 9724
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012870077530379202,
+      "loss": 0.6999,
+      "step": 9725
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012856269874273997,
+      "loss": 2.2291,
+      "step": 9726
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012842456286376642,
+      "loss": 2.3574,
+      "step": 9727
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001282863679537474,
+      "loss": 2.8599,
+      "step": 9728
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001281481142996803,
+      "loss": 3.4062,
+      "step": 9729
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012800980218868522,
+      "loss": 5.4105,
+      "step": 9730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001278714319080034,
+      "loss": 0.4816,
+      "step": 9731
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012773300374499695,
+      "loss": 2.5093,
+      "step": 9732
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012759451798714826,
+      "loss": 2.5115,
+      "step": 9733
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012745597492205937,
+      "loss": 2.9838,
+      "step": 9734
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012731737483745074,
+      "loss": 2.3719,
+      "step": 9735
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012717871802116207,
+      "loss": 2.0346,
+      "step": 9736
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012704000476115087,
+      "loss": 0.6654,
+      "step": 9737
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012690123534549083,
+      "loss": 2.7334,
+      "step": 9738
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001267624100623737,
+      "loss": 0.2164,
+      "step": 9739
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001266235292001058,
+      "loss": 0.3383,
+      "step": 9740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001264845930471103,
+      "loss": 3.5902,
+      "step": 9741
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012634560189192382,
+      "loss": 3.1147,
+      "step": 9742
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012620655602319853,
+      "loss": 3.5283,
+      "step": 9743
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001260674557296991,
+      "loss": 1.827,
+      "step": 9744
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012592830130030383,
+      "loss": 2.3629,
+      "step": 9745
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012578909302400336,
+      "loss": 2.5557,
+      "step": 9746
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012564983118990028,
+      "loss": 0.5583,
+      "step": 9747
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001255105160872077,
+      "loss": 3.5458,
+      "step": 9748
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012537114800525065,
+      "loss": 2.2262,
+      "step": 9749
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001252317272334627,
+      "loss": 3.5756,
+      "step": 9750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012509225406138826,
+      "loss": 2.7092,
+      "step": 9751
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012495272877867923,
+      "loss": 3.9503,
+      "step": 9752
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012481315167509706,
+      "loss": 2.7587,
+      "step": 9753
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012467352304050945,
+      "loss": 3.9414,
+      "step": 9754
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012453384316489253,
+      "loss": 2.8001,
+      "step": 9755
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001243941123383274,
+      "loss": 2.3231,
+      "step": 9756
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012425433085100238,
+      "loss": 1.4775,
+      "step": 9757
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012411449899320967,
+      "loss": 3.5823,
+      "step": 9758
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012397461705534739,
+      "loss": 1.8297,
+      "step": 9759
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012383468532791635,
+      "loss": 1.7602,
+      "step": 9760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012369470410152205,
+      "loss": 1.7667,
+      "step": 9761
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012355467366687167,
+      "loss": 1.7916,
+      "step": 9762
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012341459431477526,
+      "loss": 1.2789,
+      "step": 9763
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012327446633614422,
+      "loss": 3.2946,
+      "step": 9764
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012313429002199115,
+      "loss": 2.0577,
+      "step": 9765
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012299406566342836,
+      "loss": 4.3875,
+      "step": 9766
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012285379355166882,
+      "loss": 2.9703,
+      "step": 9767
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012271347397802457,
+      "loss": 0.5719,
+      "step": 9768
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001225731072339053,
+      "loss": 1.1417,
+      "step": 9769
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012243269361081997,
+      "loss": 1.2366,
+      "step": 9770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012229223340037352,
+      "loss": 1.5824,
+      "step": 9771
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012215172689426906,
+      "loss": 1.3612,
+      "step": 9772
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012201117438430442,
+      "loss": 0.7331,
+      "step": 9773
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001218705761623744,
+      "loss": 2.7503,
+      "step": 9774
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012172993252046749,
+      "loss": 1.1001,
+      "step": 9775
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001215892437506672,
+      "loss": 4.0091,
+      "step": 9776
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012144851014515058,
+      "loss": 2.5944,
+      "step": 9777
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001213077319961878,
+      "loss": 3.0367,
+      "step": 9778
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012116690959614154,
+      "loss": 2.5738,
+      "step": 9779
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012102604323746658,
+      "loss": 1.2479,
+      "step": 9780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012088513321270822,
+      "loss": 2.7307,
+      "step": 9781
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012074417981450376,
+      "loss": 2.7462,
+      "step": 9782
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012060318333557912,
+      "loss": 1.6127,
+      "step": 9783
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012046214406875107,
+      "loss": 2.8604,
+      "step": 9784
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001203210623069239,
+      "loss": 3.5576,
+      "step": 9785
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012017993834309152,
+      "loss": 3.0876,
+      "step": 9786
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00012003877247033405,
+      "loss": 2.9818,
+      "step": 9787
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011989756498182012,
+      "loss": 2.6707,
+      "step": 9788
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011975631617080333,
+      "loss": 0.931,
+      "step": 9789
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011961502633062452,
+      "loss": 2.2826,
+      "step": 9790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011947369575470831,
+      "loss": 2.6065,
+      "step": 9791
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011933232473656534,
+      "loss": 4.1761,
+      "step": 9792
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011919091356978905,
+      "loss": 4.0882,
+      "step": 9793
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001190494625480569,
+      "loss": 3.5072,
+      "step": 9794
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011890797196512892,
+      "loss": 2.7032,
+      "step": 9795
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001187664421148476,
+      "loss": 2.5943,
+      "step": 9796
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011862487329113614,
+      "loss": 3.1642,
+      "step": 9797
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011848326578799999,
+      "loss": 3.8754,
+      "step": 9798
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001183416198995235,
+      "loss": 4.1058,
+      "step": 9799
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011819993591987178,
+      "loss": 1.5176,
+      "step": 9800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011805821414328905,
+      "loss": 3.6264,
+      "step": 9801
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011791645486409699,
+      "loss": 2.2673,
+      "step": 9802
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011777465837669655,
+      "loss": 4.9674,
+      "step": 9803
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011763282497556453,
+      "loss": 4.4732,
+      "step": 9804
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011749095495525588,
+      "loss": 2.2503,
+      "step": 9805
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011734904861040026,
+      "loss": 3.0878,
+      "step": 9806
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011720710623570352,
+      "loss": 1.6565,
+      "step": 9807
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011706512812594616,
+      "loss": 2.1063,
+      "step": 9808
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011692311457598293,
+      "loss": 1.2939,
+      "step": 9809
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001167810658807421,
+      "loss": 1.578,
+      "step": 9810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011663898233522517,
+      "loss": 1.3729,
+      "step": 9811
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011649686423450528,
+      "loss": 2.2749,
+      "step": 9812
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011635471187372858,
+      "loss": 0.7951,
+      "step": 9813
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011621252554811093,
+      "loss": 0.9117,
+      "step": 9814
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001160703055529402,
+      "loss": 2.2604,
+      "step": 9815
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011592805218357272,
+      "loss": 3.6607,
+      "step": 9816
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011578576573543555,
+      "loss": 3.4337,
+      "step": 9817
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011564344650402306,
+      "loss": 0.6078,
+      "step": 9818
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00011550109478489908,
+      "loss": 2.9291,
+      "step": 9819
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001153587108736935,
+      "loss": 3.5381,
+      "step": 9820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011521629506610451,
+      "loss": 1.9217,
+      "step": 9821
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011507384765789516,
+      "loss": 2.2835,
+      "step": 9822
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011493136894489537,
+      "loss": 0.8709,
+      "step": 9823
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011478885922299906,
+      "loss": 1.3069,
+      "step": 9824
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011464631878816512,
+      "loss": 2.5964,
+      "step": 9825
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011450374793641605,
+      "loss": 1.1873,
+      "step": 9826
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001143611469638376,
+      "loss": 2.7931,
+      "step": 9827
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011421851616657795,
+      "loss": 0.3357,
+      "step": 9828
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011407585584084754,
+      "loss": 3.6243,
+      "step": 9829
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011393316628291731,
+      "loss": 3.5555,
+      "step": 9830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011379044778911982,
+      "loss": 3.1099,
+      "step": 9831
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011364770065584777,
+      "loss": 3.9125,
+      "step": 9832
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011350492517955222,
+      "loss": 2.0587,
+      "step": 9833
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011336212165674456,
+      "loss": 4.8839,
+      "step": 9834
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011321929038399298,
+      "loss": 2.8427,
+      "step": 9835
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011307643165792465,
+      "loss": 2.2177,
+      "step": 9836
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011293354577522249,
+      "loss": 1.8568,
+      "step": 9837
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011279063303262709,
+      "loss": 1.4466,
+      "step": 9838
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011264769372693367,
+      "loss": 4.086,
+      "step": 9839
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011250472815499328,
+      "loss": 1.433,
+      "step": 9840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011236173661371138,
+      "loss": 3.3384,
+      "step": 9841
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011221871940004748,
+      "loss": 2.925,
+      "step": 9842
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011207567681101378,
+      "loss": 1.5516,
+      "step": 9843
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011193260914367632,
+      "loss": 3.1683,
+      "step": 9844
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011178951669515187,
+      "loss": 3.0686,
+      "step": 9845
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001116463997626101,
+      "loss": 2.5956,
+      "step": 9846
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011150325864327,
+      "loss": 3.6475,
+      "step": 9847
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011136009363440239,
+      "loss": 2.2138,
+      "step": 9848
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011121690503332609,
+      "loss": 1.0387,
+      "step": 9849
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011107369313741048,
+      "loss": 2.9446,
+      "step": 9850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001109304582440718,
+      "loss": 1.1108,
+      "step": 9851
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011078720065077553,
+      "loss": 3.8421,
+      "step": 9852
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011064392065503289,
+      "loss": 3.1322,
+      "step": 9853
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011050061855440295,
+      "loss": 1.5865,
+      "step": 9854
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011035729464648936,
+      "loss": 2.7051,
+      "step": 9855
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011021394922894243,
+      "loss": 3.8148,
+      "step": 9856
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00011007058259945592,
+      "loss": 1.5805,
+      "step": 9857
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010992719505576836,
+      "loss": 3.5348,
+      "step": 9858
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010978378689566151,
+      "loss": 2.5002,
+      "step": 9859
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010964035841696012,
+      "loss": 0.5651,
+      "step": 9860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010949690991753044,
+      "loss": 1.8138,
+      "step": 9861
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010935344169528157,
+      "loss": 2.0618,
+      "step": 9862
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010920995404816208,
+      "loss": 3.2194,
+      "step": 9863
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010906644727416202,
+      "loss": 1.8509,
+      "step": 9864
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001089229216713111,
+      "loss": 1.7253,
+      "step": 9865
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010877937753767716,
+      "loss": 3.9273,
+      "step": 9866
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010863581517136778,
+      "loss": 1.0593,
+      "step": 9867
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010849223487052718,
+      "loss": 1.6355,
+      "step": 9868
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010834863693333808,
+      "loss": 3.7108,
+      "step": 9869
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001082050216580187,
+      "loss": 0.0703,
+      "step": 9870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010806138934282393,
+      "loss": 1.5198,
+      "step": 9871
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010791774028604384,
+      "loss": 1.1561,
+      "step": 9872
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010777407478600352,
+      "loss": 1.8111,
+      "step": 9873
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010763039314106145,
+      "loss": 2.333,
+      "step": 9874
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010748669564961085,
+      "loss": 2.9363,
+      "step": 9875
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010734298261007652,
+      "loss": 0.2174,
+      "step": 9876
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010719925432091687,
+      "loss": 2.2023,
+      "step": 9877
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010705551108062064,
+      "loss": 4.6647,
+      "step": 9878
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010691175318770901,
+      "loss": 3.0365,
+      "step": 9879
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010676798094073214,
+      "loss": 1.586,
+      "step": 9880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010662419463827148,
+      "loss": 2.3756,
+      "step": 9881
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010648039457893621,
+      "loss": 1.5683,
+      "step": 9882
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001063365810613655,
+      "loss": 2.9034,
+      "step": 9883
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010619275438422504,
+      "loss": 2.959,
+      "step": 9884
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010604891484620931,
+      "loss": 2.9504,
+      "step": 9885
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010590506274603806,
+      "loss": 0.5784,
+      "step": 9886
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010576119838245852,
+      "loss": 2.5042,
+      "step": 9887
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010561732205424219,
+      "loss": 1.1649,
+      "step": 9888
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010547343406018611,
+      "loss": 0.8072,
+      "step": 9889
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010532953469911136,
+      "loss": 1.1321,
+      "step": 9890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010518562426986278,
+      "loss": 3.6761,
+      "step": 9891
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010504170307130757,
+      "loss": 4.0937,
+      "step": 9892
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010489777140233645,
+      "loss": 1.4181,
+      "step": 9893
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010475382956186051,
+      "loss": 2.659,
+      "step": 9894
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010460987784881302,
+      "loss": 1.6637,
+      "step": 9895
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010446591656214775,
+      "loss": 1.7101,
+      "step": 9896
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010432194600083727,
+      "loss": 1.1059,
+      "step": 9897
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010417796646387487,
+      "loss": 3.0306,
+      "step": 9898
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010403397825027106,
+      "loss": 2.4616,
+      "step": 9899
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010388998165905573,
+      "loss": 2.5197,
+      "step": 9900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010374597698927497,
+      "loss": 3.906,
+      "step": 9901
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010360196453999238,
+      "loss": 2.7658,
+      "step": 9902
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010345794461028749,
+      "loss": 2.3579,
+      "step": 9903
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010331391749925538,
+      "loss": 1.5248,
+      "step": 9904
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010316988350600607,
+      "loss": 4.3293,
+      "step": 9905
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010302584292966401,
+      "loss": 1.9437,
+      "step": 9906
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010288179606936666,
+      "loss": 2.9093,
+      "step": 9907
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001027377432242657,
+      "loss": 2.6342,
+      "step": 9908
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010259368469352388,
+      "loss": 1.6965,
+      "step": 9909
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010244962077631719,
+      "loss": 2.2453,
+      "step": 9910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010230555177183134,
+      "loss": 1.4108,
+      "step": 9911
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010216147797926402,
+      "loss": 5.0766,
+      "step": 9912
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001020173996978215,
+      "loss": 3.6942,
+      "step": 9913
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010187331722672073,
+      "loss": 2.5786,
+      "step": 9914
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010172923086518599,
+      "loss": 4.0488,
+      "step": 9915
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010158514091245104,
+      "loss": 1.2679,
+      "step": 9916
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010144104766775567,
+      "loss": 1.8165,
+      "step": 9917
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010129695143034792,
+      "loss": 1.0379,
+      "step": 9918
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010115285249948084,
+      "loss": 2.9011,
+      "step": 9919
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010100875117441373,
+      "loss": 2.156,
+      "step": 9920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010086464775441074,
+      "loss": 2.9933,
+      "step": 9921
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010072054253874052,
+      "loss": 3.399,
+      "step": 9922
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010057643582667472,
+      "loss": 2.8122,
+      "step": 9923
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010043232791748942,
+      "loss": 2.8751,
+      "step": 9924
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010028821911046164,
+      "loss": 1.8008,
+      "step": 9925
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001001441097048718,
+      "loss": 2.6536,
+      "step": 9926
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00010000000000000005,
+      "loss": 1.9283,
+      "step": 9927
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.985589029512834e-05,
+      "loss": 3.5254,
+      "step": 9928
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.971178088953848e-05,
+      "loss": 2.8135,
+      "step": 9929
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.95676720825107e-05,
+      "loss": 3.282,
+      "step": 9930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.942356417332538e-05,
+      "loss": 1.1787,
+      "step": 9931
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.927945746125961e-05,
+      "loss": 1.6558,
+      "step": 9932
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.913535224558939e-05,
+      "loss": 2.3446,
+      "step": 9933
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.899124882558638e-05,
+      "loss": 3.7164,
+      "step": 9934
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.88471475005193e-05,
+      "loss": 2.3129,
+      "step": 9935
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.870304856965221e-05,
+      "loss": 1.9506,
+      "step": 9936
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.855895233224446e-05,
+      "loss": 2.084,
+      "step": 9937
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.841485908754909e-05,
+      "loss": 2.1471,
+      "step": 9938
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.827076913481413e-05,
+      "loss": 5.4263,
+      "step": 9939
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.812668277327939e-05,
+      "loss": 4.3371,
+      "step": 9940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.798260030217862e-05,
+      "loss": 3.3095,
+      "step": 9941
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.783852202073611e-05,
+      "loss": 3.3242,
+      "step": 9942
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.769444822816879e-05,
+      "loss": 1.6299,
+      "step": 9943
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.755037922368293e-05,
+      "loss": 3.6723,
+      "step": 9944
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.740631530647622e-05,
+      "loss": 2.4812,
+      "step": 9945
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.726225677573443e-05,
+      "loss": 1.77,
+      "step": 9946
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.711820393063348e-05,
+      "loss": 3.6575,
+      "step": 9947
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.69741570703361e-05,
+      "loss": 4.2527,
+      "step": 9948
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.683011649399405e-05,
+      "loss": 2.0631,
+      "step": 9949
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.668608250074474e-05,
+      "loss": 1.6673,
+      "step": 9950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.654205538971263e-05,
+      "loss": 1.0208,
+      "step": 9951
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.639803546000773e-05,
+      "loss": 2.3326,
+      "step": 9952
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.625402301072514e-05,
+      "loss": 2.0119,
+      "step": 9953
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.611001834094439e-05,
+      "loss": 4.0648,
+      "step": 9954
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.596602174972906e-05,
+      "loss": 2.3172,
+      "step": 9955
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.582203353612523e-05,
+      "loss": 2.188,
+      "step": 9956
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.567805399916284e-05,
+      "loss": 1.7039,
+      "step": 9957
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.553408343785238e-05,
+      "loss": 2.8067,
+      "step": 9958
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.53901221511871e-05,
+      "loss": 2.9664,
+      "step": 9959
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.524617043813961e-05,
+      "loss": 3.2644,
+      "step": 9960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.510222859766368e-05,
+      "loss": 1.6235,
+      "step": 9961
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.495829692869255e-05,
+      "loss": 4.1467,
+      "step": 9962
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.481437573013734e-05,
+      "loss": 2.0794,
+      "step": 9963
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.467046530088876e-05,
+      "loss": 1.4741,
+      "step": 9964
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.452656593981401e-05,
+      "loss": 0.7827,
+      "step": 9965
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.438267794575793e-05,
+      "loss": 2.6428,
+      "step": 9966
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.42388016175416e-05,
+      "loss": 2.3112,
+      "step": 9967
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.409493725396206e-05,
+      "loss": 2.6838,
+      "step": 9968
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.395108515379079e-05,
+      "loss": 1.7597,
+      "step": 9969
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.380724561577508e-05,
+      "loss": 0.2153,
+      "step": 9970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.366341893863461e-05,
+      "loss": 1.9307,
+      "step": 9971
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.35196054210639e-05,
+      "loss": 3.1741,
+      "step": 9972
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.337580536172863e-05,
+      "loss": 0.5018,
+      "step": 9973
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.323201905926796e-05,
+      "loss": 0.3933,
+      "step": 9974
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.30882468122911e-05,
+      "loss": 2.2268,
+      "step": 9975
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.294448891937947e-05,
+      "loss": 2.7534,
+      "step": 9976
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.280074567908325e-05,
+      "loss": 0.416,
+      "step": 9977
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.265701738992361e-05,
+      "loss": 2.0955,
+      "step": 9978
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.251330435038926e-05,
+      "loss": 0.9987,
+      "step": 9979
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.236960685893867e-05,
+      "loss": 1.302,
+      "step": 9980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22259252139966e-05,
+      "loss": 0.6697,
+      "step": 9981
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.208225971395626e-05,
+      "loss": 1.8609,
+      "step": 9982
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.193861065717619e-05,
+      "loss": 2.1824,
+      "step": 9983
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.179497834198141e-05,
+      "loss": 1.1455,
+      "step": 9984
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.165136306666203e-05,
+      "loss": 1.6724,
+      "step": 9985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.150776512947294e-05,
+      "loss": 3.4244,
+      "step": 9986
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.136418482863234e-05,
+      "loss": 4.3958,
+      "step": 9987
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.122062246232298e-05,
+      "loss": 1.3832,
+      "step": 9988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.1077078328689e-05,
+      "loss": 2.6992,
+      "step": 9989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09335527258381e-05,
+      "loss": 1.3594,
+      "step": 9990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.079004595183804e-05,
+      "loss": 3.1583,
+      "step": 9991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.064655830471855e-05,
+      "loss": 0.7204,
+      "step": 9992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.050309008246968e-05,
+      "loss": 2.6029,
+      "step": 9993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.035964158304e-05,
+      "loss": 1.6153,
+      "step": 9994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.021621310433861e-05,
+      "loss": 2.6631,
+      "step": 9995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.007280494423175e-05,
+      "loss": 3.0974,
+      "step": 9996
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.992941740054418e-05,
+      "loss": 2.1625,
+      "step": 9997
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.978605077105768e-05,
+      "loss": 3.2515,
+      "step": 9998
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.964270535351075e-05,
+      "loss": 0.9564,
+      "step": 9999
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.949938144559717e-05,
+      "loss": 4.2688,
+      "step": 10000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.935607934496723e-05,
+      "loss": 3.2865,
+      "step": 10001
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 8.921279934922459e-05,
+      "loss": 0.7929,
+      "step": 10002
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.906954175592832e-05,
+      "loss": 2.5113,
+      "step": 10003
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.892630686258966e-05,
+      "loss": 3.2222,
+      "step": 10004
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.878309496667403e-05,
+      "loss": 3.2905,
+      "step": 10005
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.863990636559774e-05,
+      "loss": 2.1373,
+      "step": 10006
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84967413567301e-05,
+      "loss": 5.1617,
+      "step": 10007
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.835360023739004e-05,
+      "loss": 3.8454,
+      "step": 10008
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.821048330484825e-05,
+      "loss": 2.217,
+      "step": 10009
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.806739085632381e-05,
+      "loss": 1.3807,
+      "step": 10010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.792432318898635e-05,
+      "loss": 5.3114,
+      "step": 10011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.778128059995262e-05,
+      "loss": 0.5702,
+      "step": 10012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763826338628873e-05,
+      "loss": 3.1349,
+      "step": 10013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.749527184500682e-05,
+      "loss": 4.0284,
+      "step": 10014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.735230627306644e-05,
+      "loss": 3.2191,
+      "step": 10015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.720936696737303e-05,
+      "loss": 1.8639,
+      "step": 10016
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.706645422477763e-05,
+      "loss": 4.864,
+      "step": 10017
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.692356834207546e-05,
+      "loss": 1.8491,
+      "step": 10018
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.678070961600716e-05,
+      "loss": 2.0108,
+      "step": 10019
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.663787834325555e-05,
+      "loss": 2.7284,
+      "step": 10020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.649507482044788e-05,
+      "loss": 4.2381,
+      "step": 10021
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.635229934415235e-05,
+      "loss": 3.4384,
+      "step": 10022
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.62095522108803e-05,
+      "loss": 2.3842,
+      "step": 10023
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.606683371708282e-05,
+      "loss": 1.7685,
+      "step": 10024
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.592414415915259e-05,
+      "loss": 0.9465,
+      "step": 10025
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.578148383342216e-05,
+      "loss": 4.9009,
+      "step": 10026
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.563885303616252e-05,
+      "loss": 2.3636,
+      "step": 10027
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.549625206358406e-05,
+      "loss": 2.3229,
+      "step": 10028
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.5353681211835e-05,
+      "loss": 1.2329,
+      "step": 10029
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.521114077700105e-05,
+      "loss": 2.8607,
+      "step": 10030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.506863105510474e-05,
+      "loss": 1.695,
+      "step": 10031
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.492615234210497e-05,
+      "loss": 3.7616,
+      "step": 10032
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.478370493389559e-05,
+      "loss": 1.9617,
+      "step": 10033
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.464128912630661e-05,
+      "loss": 1.4749,
+      "step": 10034
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.449890521510104e-05,
+      "loss": 2.7154,
+      "step": 10035
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.435655349597707e-05,
+      "loss": 1.9715,
+      "step": 10036
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.421423426456457e-05,
+      "loss": 1.593,
+      "step": 10037
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.40719478164274e-05,
+      "loss": 5.2119,
+      "step": 10038
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.392969444705992e-05,
+      "loss": 1.4647,
+      "step": 10039
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.378747445188918e-05,
+      "loss": 3.8777,
+      "step": 10040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.364528812627154e-05,
+      "loss": 1.4007,
+      "step": 10041
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.350313576549484e-05,
+      "loss": 1.1413,
+      "step": 10042
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.336101766477496e-05,
+      "loss": 1.6215,
+      "step": 10043
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.321893411925802e-05,
+      "loss": 1.293,
+      "step": 10044
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.307688542401719e-05,
+      "loss": 3.0207,
+      "step": 10045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.293487187405394e-05,
+      "loss": 2.3171,
+      "step": 10046
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.27928937642966e-05,
+      "loss": 3.7853,
+      "step": 10047
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.265095138959986e-05,
+      "loss": 5.1756,
+      "step": 10048
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.250904504474424e-05,
+      "loss": 1.0737,
+      "step": 10049
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.236717502443557e-05,
+      "loss": 3.4192,
+      "step": 10050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.222534162330357e-05,
+      "loss": 3.4216,
+      "step": 10051
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.208354513590313e-05,
+      "loss": 0.399,
+      "step": 10052
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.194178585671107e-05,
+      "loss": 2.7836,
+      "step": 10053
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.180006408012835e-05,
+      "loss": 2.8206,
+      "step": 10054
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.165838010047662e-05,
+      "loss": 2.4762,
+      "step": 10055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.151673421200013e-05,
+      "loss": 1.4428,
+      "step": 10056
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.137512670886398e-05,
+      "loss": 2.6175,
+      "step": 10057
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.123355788515252e-05,
+      "loss": 0.9462,
+      "step": 10058
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.109202803487119e-05,
+      "loss": 2.0489,
+      "step": 10059
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.095053745194324e-05,
+      "loss": 3.9297,
+      "step": 10060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.080908643021107e-05,
+      "loss": 1.1591,
+      "step": 10061
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.066767526343477e-05,
+      "loss": 2.1147,
+      "step": 10062
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.052630424529182e-05,
+      "loss": 1.1148,
+      "step": 10063
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.03849736693756e-05,
+      "loss": 3.047,
+      "step": 10064
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.024368382919678e-05,
+      "loss": 3.8549,
+      "step": 10065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.010243501818e-05,
+      "loss": 2.0908,
+      "step": 10066
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.996122752966606e-05,
+      "loss": 3.4733,
+      "step": 10067
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.98200616569086e-05,
+      "loss": 1.0907,
+      "step": 10068
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.967893769307621e-05,
+      "loss": 0.8156,
+      "step": 10069
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.953785593124905e-05,
+      "loss": 1.9888,
+      "step": 10070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.939681666442101e-05,
+      "loss": 3.9206,
+      "step": 10071
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.925582018549636e-05,
+      "loss": 3.1631,
+      "step": 10072
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.911486678729189e-05,
+      "loss": 4.0862,
+      "step": 10073
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.897395676253356e-05,
+      "loss": 3.86,
+      "step": 10074
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.883309040385858e-05,
+      "loss": 1.4601,
+      "step": 10075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.869226800381234e-05,
+      "loss": 0.3616,
+      "step": 10076
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.855148985484954e-05,
+      "loss": 2.4299,
+      "step": 10077
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.841075624933292e-05,
+      "loss": 4.6971,
+      "step": 10078
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.827006747953262e-05,
+      "loss": 3.3498,
+      "step": 10079
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.81294238376257e-05,
+      "loss": 0.6491,
+      "step": 10080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.798882561569568e-05,
+      "loss": 1.7112,
+      "step": 10081
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.784827310573105e-05,
+      "loss": 2.5953,
+      "step": 10082
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.77077665996266e-05,
+      "loss": 1.7628,
+      "step": 10083
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.756730638918015e-05,
+      "loss": 4.6681,
+      "step": 10084
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.742689276609482e-05,
+      "loss": 1.8905,
+      "step": 10085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.728652602197554e-05,
+      "loss": 2.1565,
+      "step": 10086
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.71462064483313e-05,
+      "loss": 4.2302,
+      "step": 10087
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.700593433657173e-05,
+      "loss": 2.0763,
+      "step": 10088
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.686570997800898e-05,
+      "loss": 2.4846,
+      "step": 10089
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.672553366385591e-05,
+      "loss": 2.6603,
+      "step": 10090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.658540568522486e-05,
+      "loss": 3.0588,
+      "step": 10091
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.644532633312842e-05,
+      "loss": 0.6371,
+      "step": 10092
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.630529589847805e-05,
+      "loss": 2.5343,
+      "step": 10093
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.616531467208377e-05,
+      "loss": 2.7854,
+      "step": 10094
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.602538294465272e-05,
+      "loss": 0.6467,
+      "step": 10095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.588550100679047e-05,
+      "loss": 3.1336,
+      "step": 10096
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.574566914899774e-05,
+      "loss": 4.8755,
+      "step": 10097
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.560588766167274e-05,
+      "loss": 2.387,
+      "step": 10098
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.54661568351076e-05,
+      "loss": 4.125,
+      "step": 10099
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.532647695949066e-05,
+      "loss": 0.5795,
+      "step": 10100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.518684832490306e-05,
+      "loss": 2.3057,
+      "step": 10101
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.504727122132089e-05,
+      "loss": 3.1966,
+      "step": 10102
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.490774593861186e-05,
+      "loss": 2.5317,
+      "step": 10103
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.476827276653742e-05,
+      "loss": 0.3961,
+      "step": 10104
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.462885199474947e-05,
+      "loss": 2.2797,
+      "step": 10105
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.44894839127924e-05,
+      "loss": 2.5034,
+      "step": 10106
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.435016881009984e-05,
+      "loss": 1.549,
+      "step": 10107
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.421090697599676e-05,
+      "loss": 1.9202,
+      "step": 10108
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.407169869969627e-05,
+      "loss": 0.6944,
+      "step": 10109
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.393254427030103e-05,
+      "loss": 3.5926,
+      "step": 10110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.379344397680159e-05,
+      "loss": 0.4131,
+      "step": 10111
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.36543981080763e-05,
+      "loss": 3.2119,
+      "step": 10112
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.35154069528898e-05,
+      "loss": 0.7217,
+      "step": 10113
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.337647079989432e-05,
+      "loss": 1.373,
+      "step": 10114
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.323758993762642e-05,
+      "loss": 4.0851,
+      "step": 10115
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.309876465450929e-05,
+      "loss": 2.6305,
+      "step": 10116
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.295999523884925e-05,
+      "loss": 0.5706,
+      "step": 10117
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.282128197883804e-05,
+      "loss": 0.6333,
+      "step": 10118
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.268262516254938e-05,
+      "loss": 2.1756,
+      "step": 10119
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.254402507794073e-05,
+      "loss": 3.0612,
+      "step": 10120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.240548201285187e-05,
+      "loss": 4.1791,
+      "step": 10121
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.226699625500317e-05,
+      "loss": 1.4046,
+      "step": 10122
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.212856809199673e-05,
+      "loss": 3.786,
+      "step": 10123
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.19901978113149e-05,
+      "loss": 1.2783,
+      "step": 10124
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.18518857003198e-05,
+      "loss": 0.6327,
+      "step": 10125
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.171363204625269e-05,
+      "loss": 2.6153,
+      "step": 10126
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.157543713623368e-05,
+      "loss": 1.9094,
+      "step": 10127
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.143730125726016e-05,
+      "loss": 2.1802,
+      "step": 10128
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.129922469620811e-05,
+      "loss": 3.1984,
+      "step": 10129
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.116120773982899e-05,
+      "loss": 4.5667,
+      "step": 10130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.102325067475182e-05,
+      "loss": 3.3107,
+      "step": 10131
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.088535378747987e-05,
+      "loss": 2.4526,
+      "step": 10132
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.074751736439283e-05,
+      "loss": 1.6568,
+      "step": 10133
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.06097416917434e-05,
+      "loss": 1.1472,
+      "step": 10134
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.047202705565952e-05,
+      "loss": 2.5909,
+      "step": 10135
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.033437374214103e-05,
+      "loss": 0.7096,
+      "step": 10136
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.019678203706173e-05,
+      "loss": 1.1343,
+      "step": 10137
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.005925222616614e-05,
+      "loss": 3.2835,
+      "step": 10138
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.99217845950716e-05,
+      "loss": 1.6388,
+      "step": 10139
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.978437942926509e-05,
+      "loss": 0.8696,
+      "step": 10140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.964703701410462e-05,
+      "loss": 3.8597,
+      "step": 10141
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.950975763481764e-05,
+      "loss": 3.5603,
+      "step": 10142
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.93725415765009e-05,
+      "loss": 0.542,
+      "step": 10143
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.923538912411898e-05,
+      "loss": 0.8109,
+      "step": 10144
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.909830056250548e-05,
+      "loss": 1.142,
+      "step": 10145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.896127617636004e-05,
+      "loss": 1.5723,
+      "step": 10146
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.882431625025034e-05,
+      "loss": 0.2743,
+      "step": 10147
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.868742106860886e-05,
+      "loss": 2.253,
+      "step": 10148
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.855059091573491e-05,
+      "loss": 2.8564,
+      "step": 10149
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.84138260757915e-05,
+      "loss": 1.0423,
+      "step": 10150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.827712683280726e-05,
+      "loss": 3.1347,
+      "step": 10151
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.814049347067331e-05,
+      "loss": 3.548,
+      "step": 10152
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.800392627314506e-05,
+      "loss": 1.2132,
+      "step": 10153
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.786742552383991e-05,
+      "loss": 2.2989,
+      "step": 10154
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.77309915062374e-05,
+      "loss": 1.5971,
+      "step": 10155
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.759462450367846e-05,
+      "loss": 4.2831,
+      "step": 10156
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.74583247993649e-05,
+      "loss": 2.8865,
+      "step": 10157
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.732209267635889e-05,
+      "loss": 0.4753,
+      "step": 10158
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.718592841758158e-05,
+      "loss": 0.2226,
+      "step": 10159
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.704983230581439e-05,
+      "loss": 1.309,
+      "step": 10160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.691380462369582e-05,
+      "loss": 1.4595,
+      "step": 10161
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.67778456537237e-05,
+      "loss": 2.754,
+      "step": 10162
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.664195567825166e-05,
+      "loss": 2.373,
+      "step": 10163
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.650613497949156e-05,
+      "loss": 0.6374,
+      "step": 10164
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.637038383950992e-05,
+      "loss": 1.4028,
+      "step": 10165
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.62347025402302e-05,
+      "loss": 0.7617,
+      "step": 10166
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.609909136342948e-05,
+      "loss": 0.8832,
+      "step": 10167
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.596355059074051e-05,
+      "loss": 1.2848,
+      "step": 10168
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.582808050364855e-05,
+      "loss": 2.7788,
+      "step": 10169
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.569268138349336e-05,
+      "loss": 1.3285,
+      "step": 10170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.555735351146615e-05,
+      "loss": 3.81,
+      "step": 10171
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.542209716861082e-05,
+      "loss": 3.8539,
+      "step": 10172
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.528691263582261e-05,
+      "loss": 2.1778,
+      "step": 10173
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.515180019384757e-05,
+      "loss": 1.2641,
+      "step": 10174
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.501676012328206e-05,
+      "loss": 0.791,
+      "step": 10175
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.48817927045723e-05,
+      "loss": 1.5608,
+      "step": 10176
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.474689821801298e-05,
+      "loss": 1.1503,
+      "step": 10177
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.461207694374846e-05,
+      "loss": 0.7611,
+      "step": 10178
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.447732916176978e-05,
+      "loss": 2.7463,
+      "step": 10179
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.43426551519166e-05,
+      "loss": 2.1494,
+      "step": 10180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.420805519387414e-05,
+      "loss": 0.8954,
+      "step": 10181
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.407352956717507e-05,
+      "loss": 2.3789,
+      "step": 10182
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.393907855119646e-05,
+      "loss": 1.7396,
+      "step": 10183
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.38047024251613e-05,
+      "loss": 1.5506,
+      "step": 10184
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.367040146813718e-05,
+      "loss": 1.1147,
+      "step": 10185
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.353617595903469e-05,
+      "loss": 1.615,
+      "step": 10186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.34020261766084e-05,
+      "loss": 1.7536,
+      "step": 10187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.326795239945545e-05,
+      "loss": 1.1254,
+      "step": 10188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.313395490601528e-05,
+      "loss": 2.6768,
+      "step": 10189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.300003397456829e-05,
+      "loss": 3.2585,
+      "step": 10190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.2866189883237e-05,
+      "loss": 2.0753,
+      "step": 10191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.273242290998307e-05,
+      "loss": 0.8815,
+      "step": 10192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.259873333260928e-05,
+      "loss": 3.0767,
+      "step": 10193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.246512142875638e-05,
+      "loss": 1.5299,
+      "step": 10194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.23315874759051e-05,
+      "loss": 2.297,
+      "step": 10195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.219813175137303e-05,
+      "loss": 2.383,
+      "step": 10196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.206475453231654e-05,
+      "loss": 2.0862,
+      "step": 10197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.193145609572771e-05,
+      "loss": 2.3722,
+      "step": 10198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.17982367184363e-05,
+      "loss": 2.0005,
+      "step": 10199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.166509667710653e-05,
+      "loss": 3.4214,
+      "step": 10200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.153203624823924e-05,
+      "loss": 1.1839,
+      "step": 10201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.139905570816857e-05,
+      "loss": 4.4515,
+      "step": 10202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.126615533306408e-05,
+      "loss": 2.6231,
+      "step": 10203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.113333539892772e-05,
+      "loss": 1.5563,
+      "step": 10204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.1000596181594996e-05,
+      "loss": 2.2278,
+      "step": 10205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.0867937956733636e-05,
+      "loss": 2.7435,
+      "step": 10206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.073536099984334e-05,
+      "loss": 2.893,
+      "step": 10207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.0602865586254386e-05,
+      "loss": 0.3556,
+      "step": 10208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.0470451991128826e-05,
+      "loss": 2.3867,
+      "step": 10209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.0338120489457506e-05,
+      "loss": 2.5843,
+      "step": 10210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.0205871356062097e-05,
+      "loss": 2.8004,
+      "step": 10211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.007370486559187e-05,
+      "loss": 1.4756,
+      "step": 10212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.994162129252581e-05,
+      "loss": 1.7144,
+      "step": 10213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.980962091116936e-05,
+      "loss": 0.6465,
+      "step": 10214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.967770399565654e-05,
+      "loss": 3.5584,
+      "step": 10215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.9545870819946704e-05,
+      "loss": 0.288,
+      "step": 10216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.9414121657826436e-05,
+      "loss": 1.7788,
+      "step": 10217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.92824567829072e-05,
+      "loss": 2.8092,
+      "step": 10218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.915087646862558e-05,
+      "loss": 1.3408,
+      "step": 10219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.90193809882427e-05,
+      "loss": 1.856,
+      "step": 10220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.888797061484286e-05,
+      "loss": 2.7696,
+      "step": 10221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.875664562133476e-05,
+      "loss": 2.2539,
+      "step": 10222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.862540628044847e-05,
+      "loss": 0.5183,
+      "step": 10223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.849425286473749e-05,
+      "loss": 3.3681,
+      "step": 10224
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.836318564657557e-05,
+      "loss": 2.7018,
+      "step": 10225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.823220489815875e-05,
+      "loss": 0.9749,
+      "step": 10226
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.810131089150219e-05,
+      "loss": 1.467,
+      "step": 10227
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.797050389844223e-05,
+      "loss": 2.8442,
+      "step": 10228
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.783978419063315e-05,
+      "loss": 3.8895,
+      "step": 10229
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.770915203954932e-05,
+      "loss": 3.7475,
+      "step": 10230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.757860771648194e-05,
+      "loss": 2.3133,
+      "step": 10231
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.7448151492541117e-05,
+      "loss": 3.4057,
+      "step": 10232
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.73177836386527e-05,
+      "loss": 0.6885,
+      "step": 10233
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.718750442556032e-05,
+      "loss": 2.3751,
+      "step": 10234
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.705731412382239e-05,
+      "loss": 1.7989,
+      "step": 10235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.692721300381331e-05,
+      "loss": 2.8312,
+      "step": 10236
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.679720133572211e-05,
+      "loss": 1.7404,
+      "step": 10237
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.6667279389552195e-05,
+      "loss": 2.6806,
+      "step": 10238
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.653744743512004e-05,
+      "loss": 1.3945,
+      "step": 10239
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.640770574205634e-05,
+      "loss": 2.4283,
+      "step": 10240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.627805457980302e-05,
+      "loss": 0.9514,
+      "step": 10241
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.61484942176153e-05,
+      "loss": 1.085,
+      "step": 10242
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.601902492455855e-05,
+      "loss": 2.9213,
+      "step": 10243
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.58896469695103e-05,
+      "loss": 2.0212,
+      "step": 10244
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.5760360621157103e-05,
+      "loss": 3.6435,
+      "step": 10245
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.563116614799653e-05,
+      "loss": 4.1611,
+      "step": 10246
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.550206381833409e-05,
+      "loss": 2.9642,
+      "step": 10247
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.5373053900285067e-05,
+      "loss": 3.7484,
+      "step": 10248
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.524413666177215e-05,
+      "loss": 1.8018,
+      "step": 10249
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.511531237052574e-05,
+      "loss": 1.1452,
+      "step": 10250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.4986581294083186e-05,
+      "loss": 2.6257,
+      "step": 10251
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.48579436997883e-05,
+      "loss": 2.862,
+      "step": 10252
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.472939985479084e-05,
+      "loss": 1.2057,
+      "step": 10253
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.460095002604529e-05,
+      "loss": 2.2397,
+      "step": 10254
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.4472594480312e-05,
+      "loss": 3.1774,
+      "step": 10255
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.434433348415421e-05,
+      "loss": 5.3744,
+      "step": 10256
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.421616730394008e-05,
+      "loss": 4.7727,
+      "step": 10257
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.4088096205839635e-05,
+      "loss": 1.3431,
+      "step": 10258
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.3960120455826633e-05,
+      "loss": 2.7605,
+      "step": 10259
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.38322403196756e-05,
+      "loss": 2.7805,
+      "step": 10260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.370445606296375e-05,
+      "loss": 3.2229,
+      "step": 10261
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.357676795106792e-05,
+      "loss": 0.6753,
+      "step": 10262
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.344917624916652e-05,
+      "loss": 5.6165,
+      "step": 10263
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.3321681222236484e-05,
+      "loss": 0.4723,
+      "step": 10264
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.3194283135055256e-05,
+      "loss": 1.1427,
+      "step": 10265
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.306698225219785e-05,
+      "loss": 0.5964,
+      "step": 10266
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.2939778838038004e-05,
+      "loss": 2.2986,
+      "step": 10267
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.281267315674694e-05,
+      "loss": 1.0713,
+      "step": 10268
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.2685665472293e-05,
+      "loss": 2.7352,
+      "step": 10269
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.255875604844045e-05,
+      "loss": 0.5435,
+      "step": 10270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.255875604844045e-05,
+      "loss": 2.9892,
+      "step": 10271
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.243194514875053e-05,
+      "loss": 3.2749,
+      "step": 10272
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.230523303657866e-05,
+      "loss": 2.292,
+      "step": 10273
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.217861997507635e-05,
+      "loss": 0.2893,
+      "step": 10274
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.205210622718813e-05,
+      "loss": 1.6637,
+      "step": 10275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.192569205565352e-05,
+      "loss": 2.0621,
+      "step": 10276
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.179937772300405e-05,
+      "loss": 4.1374,
+      "step": 10277
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.16731634915651e-05,
+      "loss": 0.6077,
+      "step": 10278
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.154704962345295e-05,
+      "loss": 1.7658,
+      "step": 10279
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.142103638057667e-05,
+      "loss": 0.799,
+      "step": 10280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.1295124024635156e-05,
+      "loss": 2.9869,
+      "step": 10281
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.116931281711881e-05,
+      "loss": 3.1025,
+      "step": 10282
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.104360301930739e-05,
+      "loss": 4.9373,
+      "step": 10283
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.0917994892270225e-05,
+      "loss": 2.2229,
+      "step": 10284
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.0792488696865606e-05,
+      "loss": 1.025,
+      "step": 10285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.066708469373953e-05,
+      "loss": 1.4987,
+      "step": 10286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.054178314332687e-05,
+      "loss": 4.4336,
+      "step": 10287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.041658430584849e-05,
+      "loss": 0.963,
+      "step": 10288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.029148844131315e-05,
+      "loss": 2.9254,
+      "step": 10289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.0166495809514546e-05,
+      "loss": 1.8867,
+      "step": 10290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 5.004160667003322e-05,
+      "loss": 2.1036,
+      "step": 10291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.991682128223354e-05,
+      "loss": 2.2623,
+      "step": 10292
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.979213990526567e-05,
+      "loss": 1.3128,
+      "step": 10293
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.966756279806247e-05,
+      "loss": 1.0485,
+      "step": 10294
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.954309021934158e-05,
+      "loss": 2.4088,
+      "step": 10295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.941872242760225e-05,
+      "loss": 0.505,
+      "step": 10296
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.929445968112737e-05,
+      "loss": 3.4914,
+      "step": 10297
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.9170302237980606e-05,
+      "loss": 1.1663,
+      "step": 10298
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.904625035600755e-05,
+      "loss": 1.996,
+      "step": 10299
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.892230429283441e-05,
+      "loss": 3.1868,
+      "step": 10300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.879846430586763e-05,
+      "loss": 3.2982,
+      "step": 10301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.867473065229336e-05,
+      "loss": 2.3826,
+      "step": 10302
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.8551103589077084e-05,
+      "loss": 5.7259,
+      "step": 10303
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.842758337296229e-05,
+      "loss": 2.138,
+      "step": 10304
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.830417026047167e-05,
+      "loss": 0.1008,
+      "step": 10305
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.818086450790423e-05,
+      "loss": 1.9958,
+      "step": 10306
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.8057666371337265e-05,
+      "loss": 4.8281,
+      "step": 10307
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.793457610662335e-05,
+      "loss": 3.0973,
+      "step": 10308
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.7811593969392246e-05,
+      "loss": 0.6659,
+      "step": 10309
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.768872021504794e-05,
+      "loss": 2.0397,
+      "step": 10310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.7565955098770576e-05,
+      "loss": 2.042,
+      "step": 10311
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.744329887551343e-05,
+      "loss": 3.5946,
+      "step": 10312
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.7320751800004725e-05,
+      "loss": 3.975,
+      "step": 10313
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.719831412674539e-05,
+      "loss": 0.8447,
+      "step": 10314
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.707598611000931e-05,
+      "loss": 3.0616,
+      "step": 10315
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.695376800384278e-05,
+      "loss": 4.6284,
+      "step": 10316
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.6831660062063234e-05,
+      "loss": 1.5099,
+      "step": 10317
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.670966253826038e-05,
+      "loss": 1.4653,
+      "step": 10318
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.6587775685793436e-05,
+      "loss": 1.3463,
+      "step": 10319
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.646599975779292e-05,
+      "loss": 2.9543,
+      "step": 10320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.634433500715787e-05,
+      "loss": 2.9088,
+      "step": 10321
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.6222781686557557e-05,
+      "loss": 0.3228,
+      "step": 10322
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.6101340048428676e-05,
+      "loss": 3.354,
+      "step": 10323
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.59800103449772e-05,
+      "loss": 2.9431,
+      "step": 10324
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.58587928281754e-05,
+      "loss": 2.2641,
+      "step": 10325
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.573768774976379e-05,
+      "loss": 4.0096,
+      "step": 10326
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.561669536124816e-05,
+      "loss": 1.3314,
+      "step": 10327
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5495815913901496e-05,
+      "loss": 0.7573,
+      "step": 10328
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.5375049658760995e-05,
+      "loss": 2.2467,
+      "step": 10329
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.525439684663001e-05,
+      "loss": 1.5226,
+      "step": 10330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.513385772807525e-05,
+      "loss": 3.9294,
+      "step": 10331
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.501343255342788e-05,
+      "loss": 2.0436,
+      "step": 10332
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.48931215727823e-05,
+      "loss": 1.4714,
+      "step": 10333
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.477292503599593e-05,
+      "loss": 1.7926,
+      "step": 10334
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.465284319268788e-05,
+      "loss": 2.5532,
+      "step": 10335
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4532876292240125e-05,
+      "loss": 2.5592,
+      "step": 10336
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.441302458379475e-05,
+      "loss": 5.2924,
+      "step": 10337
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.4293288316255796e-05,
+      "loss": 0.1836,
+      "step": 10338
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.417366773828638e-05,
+      "loss": 0.6974,
+      "step": 10339
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.405416309831053e-05,
+      "loss": 2.552,
+      "step": 10340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3934774644510324e-05,
+      "loss": 1.6369,
+      "step": 10341
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3815502624827745e-05,
+      "loss": 2.8233,
+      "step": 10342
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3696347286961744e-05,
+      "loss": 1.4021,
+      "step": 10343
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3577308878370194e-05,
+      "loss": 0.2775,
+      "step": 10344
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.345838764626691e-05,
+      "loss": 3.4261,
+      "step": 10345
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.333958383762341e-05,
+      "loss": 2.4913,
+      "step": 10346
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.322089769916688e-05,
+      "loss": 2.1415,
+      "step": 10347
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.3102329477379754e-05,
+      "loss": 3.9243,
+      "step": 10348
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.298387941850054e-05,
+      "loss": 2.891,
+      "step": 10349
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.286554776852122e-05,
+      "loss": 5.4936,
+      "step": 10350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.274733477318903e-05,
+      "loss": 3.5637,
+      "step": 10351
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.2629240678003604e-05,
+      "loss": 3.0076,
+      "step": 10352
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.251126572821884e-05,
+      "loss": 2.2713,
+      "step": 10353
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.239341016884002e-05,
+      "loss": 1.5377,
+      "step": 10354
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.227567424462565e-05,
+      "loss": 0.8237,
+      "step": 10355
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.21580582000846e-05,
+      "loss": 0.2634,
+      "step": 10356
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.204056227947798e-05,
+      "loss": 2.1486,
+      "step": 10357
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1923186726816245e-05,
+      "loss": 1.4876,
+      "step": 10358
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.180593178586102e-05,
+      "loss": 4.7726,
+      "step": 10359
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1688797700122304e-05,
+      "loss": 1.776,
+      "step": 10360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1571784712860296e-05,
+      "loss": 0.831,
+      "step": 10361
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1454893067082634e-05,
+      "loss": 0.955,
+      "step": 10362
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.133812300554558e-05,
+      "loss": 2.6394,
+      "step": 10363
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.122147477075273e-05,
+      "loss": 2.6214,
+      "step": 10364
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.110494860495483e-05,
+      "loss": 0.7648,
+      "step": 10365
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.098854475014853e-05,
+      "loss": 1.744,
+      "step": 10366
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0872263448077485e-05,
+      "loss": 2.6258,
+      "step": 10367
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.075610494022968e-05,
+      "loss": 0.6115,
+      "step": 10368
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.064006946783925e-05,
+      "loss": 0.3783,
+      "step": 10369
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0524157271883634e-05,
+      "loss": 2.4672,
+      "step": 10370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0408368593085446e-05,
+      "loss": 1.8206,
+      "step": 10371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.0292703671909607e-05,
+      "loss": 3.2971,
+      "step": 10372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.017716274856518e-05,
+      "loss": 2.896,
+      "step": 10373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.006174606300256e-05,
+      "loss": 1.2538,
+      "step": 10374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.994645385491528e-05,
+      "loss": 1.209,
+      "step": 10375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.983128636373723e-05,
+      "loss": 2.2192,
+      "step": 10376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9716243828644284e-05,
+      "loss": 1.4042,
+      "step": 10377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.960132648855223e-05,
+      "loss": 0.7395,
+      "step": 10378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9486534582117005e-05,
+      "loss": 0.6671,
+      "step": 10379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.937186834773421e-05,
+      "loss": 0.8925,
+      "step": 10380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.925732802353785e-05,
+      "loss": 1.7225,
+      "step": 10381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.91429138474015e-05,
+      "loss": 0.6409,
+      "step": 10382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9028626056935534e-05,
+      "loss": 0.4398,
+      "step": 10383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8914464889489054e-05,
+      "loss": 2.2547,
+      "step": 10384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.880043058214705e-05,
+      "loss": 1.0914,
+      "step": 10385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.868652337173217e-05,
+      "loss": 1.579,
+      "step": 10386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.857274349480201e-05,
+      "loss": 1.8238,
+      "step": 10387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8459091187650806e-05,
+      "loss": 2.6154,
+      "step": 10388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.834556668630678e-05,
+      "loss": 1.2687,
+      "step": 10389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8232170226533826e-05,
+      "loss": 0.6869,
+      "step": 10390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.811890204382883e-05,
+      "loss": 2.276,
+      "step": 10391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.800576237342338e-05,
+      "loss": 1.9106,
+      "step": 10392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.789275145028121e-05,
+      "loss": 1.5786,
+      "step": 10393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.7779869509099194e-05,
+      "loss": 2.9523,
+      "step": 10394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.7667116784306276e-05,
+      "loss": 3.4921,
+      "step": 10395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.755449351006298e-05,
+      "loss": 1.9918,
+      "step": 10396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.7441999920261015e-05,
+      "loss": 3.9069,
+      "step": 10397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.7329636248522924e-05,
+      "loss": 2.1178,
+      "step": 10398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.7217402728200865e-05,
+      "loss": 2.1556,
+      "step": 10399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.710529959237767e-05,
+      "loss": 2.001,
+      "step": 10400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.6993327073864306e-05,
+      "loss": 0.1924,
+      "step": 10401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.6881485405201556e-05,
+      "loss": 0.2184,
+      "step": 10402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.6769774818657374e-05,
+      "loss": 1.0716,
+      "step": 10403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.665819554622861e-05,
+      "loss": 0.7518,
+      "step": 10404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.654674781963826e-05,
+      "loss": 0.4841,
+      "step": 10405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.643543187033727e-05,
+      "loss": 3.9312,
+      "step": 10406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.632424792950178e-05,
+      "loss": 1.3098,
+      "step": 10407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.621319622803489e-05,
+      "loss": 1.6852,
+      "step": 10408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.610227699656402e-05,
+      "loss": 4.1276,
+      "step": 10409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.599149046544238e-05,
+      "loss": 3.3207,
+      "step": 10410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.58808368647472e-05,
+      "loss": 0.6929,
+      "step": 10411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.5770316424279246e-05,
+      "loss": 2.3978,
+      "step": 10412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.5659929373563695e-05,
+      "loss": 2.2675,
+      "step": 10413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.5549675941847594e-05,
+      "loss": 3.6903,
+      "step": 10414
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.543955635810161e-05,
+      "loss": 2.579,
+      "step": 10415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.5329570851017335e-05,
+      "loss": 0.4097,
+      "step": 10416
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.5219719649009034e-05,
+      "loss": 1.0656,
+      "step": 10417
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.511000298021091e-05,
+      "loss": 1.8162,
+      "step": 10418
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.500042107247892e-05,
+      "loss": 0.6479,
+      "step": 10419
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.4890974153388015e-05,
+      "loss": 1.4344,
+      "step": 10420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.478166245023392e-05,
+      "loss": 0.484,
+      "step": 10421
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.467248619003047e-05,
+      "loss": 3.3425,
+      "step": 10422
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.456344559951129e-05,
+      "loss": 2.6573,
+      "step": 10423
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.445454090512731e-05,
+      "loss": 1.2282,
+      "step": 10424
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.434577233304775e-05,
+      "loss": 1.0438,
+      "step": 10425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.423714010915906e-05,
+      "loss": 4.4102,
+      "step": 10426
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.412864445906445e-05,
+      "loss": 3.9699,
+      "step": 10427
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.4020285608083566e-05,
+      "loss": 1.8797,
+      "step": 10428
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.391206378125206e-05,
+      "loss": 2.9364,
+      "step": 10429
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.380397920332048e-05,
+      "loss": 0.6934,
+      "step": 10430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.369603209875528e-05,
+      "loss": 0.8425,
+      "step": 10431
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.3588222691736346e-05,
+      "loss": 2.3253,
+      "step": 10432
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.348055120615867e-05,
+      "loss": 1.4101,
+      "step": 10433
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.337301786562975e-05,
+      "loss": 1.5319,
+      "step": 10434
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.326562289347124e-05,
+      "loss": 2.2348,
+      "step": 10435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.31583665127164e-05,
+      "loss": 2.8052,
+      "step": 10436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.305124894611169e-05,
+      "loss": 0.6668,
+      "step": 10437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.2944270416114267e-05,
+      "loss": 1.6167,
+      "step": 10438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.2837431144893574e-05,
+      "loss": 3.3091,
+      "step": 10439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.273073135432879e-05,
+      "loss": 0.9406,
+      "step": 10440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.262417126601037e-05,
+      "loss": 4.6162,
+      "step": 10441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.251775110123825e-05,
+      "loss": 1.3941,
+      "step": 10442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.2411471081021324e-05,
+      "loss": 2.4996,
+      "step": 10443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.230533142607843e-05,
+      "loss": 0.7087,
+      "step": 10444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.219933235683582e-05,
+      "loss": 0.7168,
+      "step": 10445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.2093474093428834e-05,
+      "loss": 4.25,
+      "step": 10446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1987756855699346e-05,
+      "loss": 2.8765,
+      "step": 10447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1882180863197375e-05,
+      "loss": 2.5657,
+      "step": 10448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.177674633517862e-05,
+      "loss": 2.7943,
+      "step": 10449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.167145349060597e-05,
+      "loss": 1.2943,
+      "step": 10450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.156630254814709e-05,
+      "loss": 3.5936,
+      "step": 10451
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.146129372617593e-05,
+      "loss": 2.746,
+      "step": 10452
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.135642724277029e-05,
+      "loss": 3.9425,
+      "step": 10453
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.12517033157134e-05,
+      "loss": 3.3841,
+      "step": 10454
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.114712216249139e-05,
+      "loss": 2.753,
+      "step": 10455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1042684000294906e-05,
+      "loss": 1.6987,
+      "step": 10456
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.093838904601676e-05,
+      "loss": 1.5814,
+      "step": 10457
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.083423751625285e-05,
+      "loss": 1.3745,
+      "step": 10458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.07302296273011e-05,
+      "loss": 0.275,
+      "step": 10459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.062636559516122e-05,
+      "loss": 0.4405,
+      "step": 10460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.052264563553372e-05,
+      "loss": 4.3807,
+      "step": 10461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.0419069963820722e-05,
+      "loss": 3.6546,
+      "step": 10462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.0315638795123735e-05,
+      "loss": 2.4353,
+      "step": 10463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.0212352344245155e-05,
+      "loss": 3.7547,
+      "step": 10464
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.010921082568583e-05,
+      "loss": 2.2513,
+      "step": 10465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.000621445364664e-05,
+      "loss": 3.2066,
+      "step": 10466
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9903363442026e-05,
+      "loss": 3.3915,
+      "step": 10467
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9800658004421467e-05,
+      "loss": 0.9224,
+      "step": 10468
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9698098354127257e-05,
+      "loss": 1.4075,
+      "step": 10469
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.959568470413585e-05,
+      "loss": 1.3031,
+      "step": 10470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.949341726713548e-05,
+      "loss": 2.7303,
+      "step": 10471
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9391296255511768e-05,
+      "loss": 3.2014,
+      "step": 10472
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9289321881345356e-05,
+      "loss": 1.0902,
+      "step": 10473
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.9187494356412702e-05,
+      "loss": 2.6054,
+      "step": 10474
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.908581389218573e-05,
+      "loss": 0.4193,
+      "step": 10475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8984280699830003e-05,
+      "loss": 0.9716,
+      "step": 10476
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.888289499020631e-05,
+      "loss": 1.6984,
+      "step": 10477
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8781656973868087e-05,
+      "loss": 1.3268,
+      "step": 10478
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8680566861063097e-05,
+      "loss": 2.1257,
+      "step": 10479
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8579624861730926e-05,
+      "loss": 3.0217,
+      "step": 10480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.847883118550454e-05,
+      "loss": 2.3456,
+      "step": 10481
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.837818604170791e-05,
+      "loss": 2.1674,
+      "step": 10482
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.827768963935752e-05,
+      "loss": 2.2049,
+      "step": 10483
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.817734218715996e-05,
+      "loss": 1.5781,
+      "step": 10484
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.807714389351348e-05,
+      "loss": 0.1752,
+      "step": 10485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.797709496650557e-05,
+      "loss": 1.8354,
+      "step": 10486
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.787719561391451e-05,
+      "loss": 2.6711,
+      "step": 10487
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.7777446043207088e-05,
+      "loss": 2.0319,
+      "step": 10488
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.7677846461539525e-05,
+      "loss": 1.2688,
+      "step": 10489
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.7578397075756433e-05,
+      "loss": 2.0029,
+      "step": 10490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.747909809239061e-05,
+      "loss": 2.942,
+      "step": 10491
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.7379949717662023e-05,
+      "loss": 2.605,
+      "step": 10492
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.7280952157478713e-05,
+      "loss": 2.789,
+      "step": 10493
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.718210561743455e-05,
+      "loss": 4.111,
+      "step": 10494
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.708341030281073e-05,
+      "loss": 1.1456,
+      "step": 10495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.698486641857343e-05,
+      "loss": 1.6698,
+      "step": 10496
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.688647416937531e-05,
+      "loss": 3.4488,
+      "step": 10497
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.678823375955314e-05,
+      "loss": 0.327,
+      "step": 10498
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.669014539312934e-05,
+      "loss": 4.249,
+      "step": 10499
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.6592209273809588e-05,
+      "loss": 2.0163,
+      "step": 10500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.649442560498436e-05,
+      "loss": 2.0306,
+      "step": 10501
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.639679458972657e-05,
+      "loss": 2.0049,
+      "step": 10502
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.6299316430793054e-05,
+      "loss": 1.6334,
+      "step": 10503
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.6201991330622234e-05,
+      "loss": 1.2689,
+      "step": 10504
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.6104819491335664e-05,
+      "loss": 0.9598,
+      "step": 10505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.600780111473572e-05,
+      "loss": 1.9062,
+      "step": 10506
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.5910936402306475e-05,
+      "loss": 1.3236,
+      "step": 10507
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.5814225555213223e-05,
+      "loss": 2.8897,
+      "step": 10508
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.5717668774300863e-05,
+      "loss": 0.8614,
+      "step": 10509
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.5621266260095322e-05,
+      "loss": 2.4171,
+      "step": 10510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.5525018212801168e-05,
+      "loss": 2.1744,
+      "step": 10511
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.5428924832303135e-05,
+      "loss": 1.149,
+      "step": 10512
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.533298631816383e-05,
+      "loss": 1.1236,
+      "step": 10513
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.5237202869625154e-05,
+      "loss": 2.3704,
+      "step": 10514
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.514157468560605e-05,
+      "loss": 0.9094,
+      "step": 10515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.5046101964703926e-05,
+      "loss": 2.9539,
+      "step": 10516
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.4950784905192426e-05,
+      "loss": 0.2934,
+      "step": 10517
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.4855623705022845e-05,
+      "loss": 0.5453,
+      "step": 10518
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.476061856182197e-05,
+      "loss": 1.0858,
+      "step": 10519
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.466576967289298e-05,
+      "loss": 2.2306,
+      "step": 10520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.45710772352144e-05,
+      "loss": 0.8011,
+      "step": 10521
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.4476541445439893e-05,
+      "loss": 1.1091,
+      "step": 10522
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.4382162499897742e-05,
+      "loss": 2.6393,
+      "step": 10523
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.4287940594590686e-05,
+      "loss": 2.6741,
+      "step": 10524
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.419387592519481e-05,
+      "loss": 2.3711,
+      "step": 10525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.409996868706048e-05,
+      "loss": 3.0952,
+      "step": 10526
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.4006219075210213e-05,
+      "loss": 0.1267,
+      "step": 10527
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.3912627284340073e-05,
+      "loss": 1.3631,
+      "step": 10528
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.381919350881745e-05,
+      "loss": 1.9465,
+      "step": 10529
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.372591794268251e-05,
+      "loss": 1.5614,
+      "step": 10530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.3632800779645915e-05,
+      "loss": 0.5182,
+      "step": 10531
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.353984221309029e-05,
+      "loss": 4.3837,
+      "step": 10532
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.344704243606799e-05,
+      "loss": 0.1937,
+      "step": 10533
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.3354401641302482e-05,
+      "loss": 2.1889,
+      "step": 10534
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.3261920021186178e-05,
+      "loss": 3.6175,
+      "step": 10535
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.3169597767781825e-05,
+      "loss": 4.1136,
+      "step": 10536
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.3077435072820418e-05,
+      "loss": 1.6615,
+      "step": 10537
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.2985432127701922e-05,
+      "loss": 1.2807,
+      "step": 10538
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.2893589123494873e-05,
+      "loss": 2.5157,
+      "step": 10539
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.2801906250934845e-05,
+      "loss": 1.6988,
+      "step": 10540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.2710383700425797e-05,
+      "loss": 1.1753,
+      "step": 10541
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.26190216620378e-05,
+      "loss": 2.6184,
+      "step": 10542
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.2527820325508496e-05,
+      "loss": 0.2432,
+      "step": 10543
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.24367798802409e-05,
+      "loss": 2.8953,
+      "step": 10544
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.234590051530475e-05,
+      "loss": 0.3839,
+      "step": 10545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.2255182419434416e-05,
+      "loss": 2.5118,
+      "step": 10546
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.21646257810302e-05,
+      "loss": 1.0877,
+      "step": 10547
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.2074230788156238e-05,
+      "loss": 0.6237,
+      "step": 10548
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1983997628541832e-05,
+      "loss": 1.6746,
+      "step": 10549
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.189392648957943e-05,
+      "loss": 3.747,
+      "step": 10550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.180401755832544e-05,
+      "loss": 1.4494,
+      "step": 10551
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.17142710214993e-05,
+      "loss": 2.9223,
+      "step": 10552
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.1624687065483173e-05,
+      "loss": 2.2262,
+      "step": 10553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.153526587632161e-05,
+      "loss": 3.9209,
+      "step": 10554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.144600763972122e-05,
+      "loss": 2.1891,
+      "step": 10555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1356912541049778e-05,
+      "loss": 5.7566,
+      "step": 10556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1267980765337005e-05,
+      "loss": 3.1726,
+      "step": 10557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1179212497272594e-05,
+      "loss": 2.5233,
+      "step": 10558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1090607921207507e-05,
+      "loss": 1.7245,
+      "step": 10559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.1002167221151924e-05,
+      "loss": 1.9202,
+      "step": 10560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.091389058077653e-05,
+      "loss": 5.073,
+      "step": 10561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.082577818341046e-05,
+      "loss": 0.5368,
+      "step": 10562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0737830212042598e-05,
+      "loss": 2.8783,
+      "step": 10563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0650046849319492e-05,
+      "loss": 1.0303,
+      "step": 10564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0562428277546718e-05,
+      "loss": 3.8546,
+      "step": 10565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0474974678686743e-05,
+      "loss": 2.6037,
+      "step": 10566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.038768623436028e-05,
+      "loss": 0.3749,
+      "step": 10567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0300563125844317e-05,
+      "loss": 0.3612,
+      "step": 10568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.021360553407292e-05,
+      "loss": 0.7583,
+      "step": 10569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0126813639636267e-05,
+      "loss": 0.3993,
+      "step": 10570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0040187622780347e-05,
+      "loss": 2.1225,
+      "step": 10571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9953727663407096e-05,
+      "loss": 2.7754,
+      "step": 10572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.986743394107281e-05,
+      "loss": 2.533,
+      "step": 10573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9781306634989426e-05,
+      "loss": 1.5982,
+      "step": 10574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9695345924022412e-05,
+      "loss": 0.1662,
+      "step": 10575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.960955198669212e-05,
+      "loss": 3.6552,
+      "step": 10576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9523925001171693e-05,
+      "loss": 4.6802,
+      "step": 10577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.943846514528841e-05,
+      "loss": 1.2589,
+      "step": 10578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9353172596521598e-05,
+      "loss": 1.3272,
+      "step": 10579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9268047532003997e-05,
+      "loss": 2.7829,
+      "step": 10580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9183090128519654e-05,
+      "loss": 1.8829,
+      "step": 10581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.909830056250529e-05,
+      "loss": 2.3292,
+      "step": 10582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9013679010048325e-05,
+      "loss": 0.378,
+      "step": 10583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.892922564688765e-05,
+      "loss": 1.1631,
+      "step": 10584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.884494064841279e-05,
+      "loss": 0.7697,
+      "step": 10585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.876082418966372e-05,
+      "loss": 1.4932,
+      "step": 10586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8676876445329973e-05,
+      "loss": 1.0012,
+      "step": 10587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8593097589751428e-05,
+      "loss": 1.3025,
+      "step": 10588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.850948779691639e-05,
+      "loss": 1.3859,
+      "step": 10589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8426047240462907e-05,
+      "loss": 2.5442,
+      "step": 10590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.834277609367673e-05,
+      "loss": 4.7541,
+      "step": 10591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8259674529492577e-05,
+      "loss": 4.6147,
+      "step": 10592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8176742720492225e-05,
+      "loss": 2.3142,
+      "step": 10593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8093980838905668e-05,
+      "loss": 2.0573,
+      "step": 10594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.801138905660922e-05,
+      "loss": 0.75,
+      "step": 10595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7928967545126753e-05,
+      "loss": 1.6258,
+      "step": 10596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.784671647562769e-05,
+      "loss": 0.9664,
+      "step": 10597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7764636018928328e-05,
+      "loss": 3.216,
+      "step": 10598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7682726345489818e-05,
+      "loss": 5.3969,
+      "step": 10599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.760098762541943e-05,
+      "loss": 0.1958,
+      "step": 10600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7519420028468725e-05,
+      "loss": 1.2889,
+      "step": 10601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.743802372403416e-05,
+      "loss": 1.2252,
+      "step": 10602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7356798881156778e-05,
+      "loss": 2.0311,
+      "step": 10603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7275745668520826e-05,
+      "loss": 4.5561,
+      "step": 10604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7194864254454923e-05,
+      "loss": 1.2686,
+      "step": 10605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.7114154806930095e-05,
+      "loss": 2.0299,
+      "step": 10606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.703361749356103e-05,
+      "loss": 1.3506,
+      "step": 10607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6953252481604155e-05,
+      "loss": 4.6206,
+      "step": 10608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6873059937958845e-05,
+      "loss": 2.4428,
+      "step": 10609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.679304002916552e-05,
+      "loss": 2.9099,
+      "step": 10610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.671319292140686e-05,
+      "loss": 0.6286,
+      "step": 10611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6633518780505887e-05,
+      "loss": 2.9574,
+      "step": 10612
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.655401777192722e-05,
+      "loss": 2.6597,
+      "step": 10613
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.647469006077521e-05,
+      "loss": 0.8124,
+      "step": 10614
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6395535811794715e-05,
+      "loss": 1.2733,
+      "step": 10615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6316555189370254e-05,
+      "loss": 2.7196,
+      "step": 10616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.623774835752586e-05,
+      "loss": 1.8772,
+      "step": 10617
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6159115479924257e-05,
+      "loss": 0.7677,
+      "step": 10618
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6080656719867603e-05,
+      "loss": 0.9752,
+      "step": 10619
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.600237224029566e-05,
+      "loss": 3.3401,
+      "step": 10620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5924262203787012e-05,
+      "loss": 3.0225,
+      "step": 10621
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.584632677255723e-05,
+      "loss": 2.744,
+      "step": 10622
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5768566108460015e-05,
+      "loss": 2.245,
+      "step": 10623
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5690980372985366e-05,
+      "loss": 0.2055,
+      "step": 10624
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5613569727260745e-05,
+      "loss": 4.2659,
+      "step": 10625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.553633433204925e-05,
+      "loss": 2.1619,
+      "step": 10626
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5459274347750774e-05,
+      "loss": 1.7173,
+      "step": 10627
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5382389934400178e-05,
+      "loss": 1.432,
+      "step": 10628
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5305681251668437e-05,
+      "loss": 0.8657,
+      "step": 10629
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5229148458860842e-05,
+      "loss": 2.0373,
+      "step": 10630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5152791714918158e-05,
+      "loss": 1.8837,
+      "step": 10631
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5076611178414845e-05,
+      "loss": 1.4813,
+      "step": 10632
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.5000607007559841e-05,
+      "loss": 1.4262,
+      "step": 10633
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4924779360195706e-05,
+      "loss": 0.8947,
+      "step": 10634
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4849128393798306e-05,
+      "loss": 2.2413,
+      "step": 10635
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4773654265476944e-05,
+      "loss": 2.4943,
+      "step": 10636
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4698357131973051e-05,
+      "loss": 2.9696,
+      "step": 10637
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4623237149661195e-05,
+      "loss": 1.9031,
+      "step": 10638
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4548294474547308e-05,
+      "loss": 1.6972,
+      "step": 10639
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4473529262269858e-05,
+      "loss": 2.9289,
+      "step": 10640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4398941668098009e-05,
+      "loss": 1.3396,
+      "step": 10641
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4324531846932787e-05,
+      "loss": 0.9052,
+      "step": 10642
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4250299953305301e-05,
+      "loss": 0.727,
+      "step": 10643
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.417624614137788e-05,
+      "loss": 3.7357,
+      "step": 10644
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4102370564942368e-05,
+      "loss": 1.6169,
+      "step": 10645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4028673377420842e-05,
+      "loss": 3.7887,
+      "step": 10646
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3955154731864795e-05,
+      "loss": 4.1344,
+      "step": 10647
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.388181478095495e-05,
+      "loss": 2.4006,
+      "step": 10648
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3808653677000894e-05,
+      "loss": 2.7653,
+      "step": 10649
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3735671571940922e-05,
+      "loss": 0.2199,
+      "step": 10650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3662868617341207e-05,
+      "loss": 1.2593,
+      "step": 10651
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3590244964396492e-05,
+      "loss": 3.5337,
+      "step": 10652
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3517800763928457e-05,
+      "loss": 1.5516,
+      "step": 10653
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3445536166386807e-05,
+      "loss": 2.1226,
+      "step": 10654
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.337345132184753e-05,
+      "loss": 0.1367,
+      "step": 10655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3301546380014029e-05,
+      "loss": 2.6639,
+      "step": 10656
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.32298214902154e-05,
+      "loss": 2.4587,
+      "step": 10657
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3158276801407499e-05,
+      "loss": 2.946,
+      "step": 10658
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.3086912462171242e-05,
+      "loss": 0.406,
+      "step": 10659
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.301572862071373e-05,
+      "loss": 1.2534,
+      "step": 10660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2944725424866466e-05,
+      "loss": 3.3144,
+      "step": 10661
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2873903022086554e-05,
+      "loss": 0.1327,
+      "step": 10662
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2803261559454972e-05,
+      "loss": 1.8551,
+      "step": 10663
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2732801183677311e-05,
+      "loss": 1.4374,
+      "step": 10664
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.266252204108297e-05,
+      "loss": 0.2715,
+      "step": 10665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2592424277624915e-05,
+      "loss": 3.0974,
+      "step": 10666
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2522508038879743e-05,
+      "loss": 2.726,
+      "step": 10667
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2452773470046509e-05,
+      "loss": 1.4198,
+      "step": 10668
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2383220715947619e-05,
+      "loss": 0.6299,
+      "step": 10669
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2313849921027243e-05,
+      "loss": 1.0098,
+      "step": 10670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2244661229352339e-05,
+      "loss": 2.5944,
+      "step": 10671
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.217565478461099e-05,
+      "loss": 0.2715,
+      "step": 10672
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2106830730113484e-05,
+      "loss": 2.7226,
+      "step": 10673
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.2038189208790663e-05,
+      "loss": 0.6656,
+      "step": 10674
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1969730363194931e-05,
+      "loss": 2.7375,
+      "step": 10675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1901454335498685e-05,
+      "loss": 1.9223,
+      "step": 10676
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1833361267495303e-05,
+      "loss": 2.9574,
+      "step": 10677
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1765451300597597e-05,
+      "loss": 0.7696,
+      "step": 10678
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1697724575838487e-05,
+      "loss": 2.1811,
+      "step": 10679
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1630181233870253e-05,
+      "loss": 3.6627,
+      "step": 10680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1562821414964398e-05,
+      "loss": 3.5938,
+      "step": 10681
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1495645259010968e-05,
+      "loss": 1.0437,
+      "step": 10682
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1428652905519188e-05,
+      "loss": 3.9825,
+      "step": 10683
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1361844493615859e-05,
+      "loss": 0.6408,
+      "step": 10684
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1295220162046494e-05,
+      "loss": 2.5445,
+      "step": 10685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1228780049173615e-05,
+      "loss": 2.5028,
+      "step": 10686
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1162524292977838e-05,
+      "loss": 1.4397,
+      "step": 10687
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1096453031056264e-05,
+      "loss": 2.4183,
+      "step": 10688
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.1030566400623476e-05,
+      "loss": 1.9701,
+      "step": 10689
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0964864538510011e-05,
+      "loss": 1.425,
+      "step": 10690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0899347581163277e-05,
+      "loss": 3.3313,
+      "step": 10691
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.083401566464609e-05,
+      "loss": 2.7719,
+      "step": 10692
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0768868924637598e-05,
+      "loss": 2.8724,
+      "step": 10693
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0703907496431797e-05,
+      "loss": 1.4147,
+      "step": 10694
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0639131514938206e-05,
+      "loss": 3.5896,
+      "step": 10695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.057454111468109e-05,
+      "loss": 2.2173,
+      "step": 10696
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0510136429799333e-05,
+      "loss": 1.9471,
+      "step": 10697
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0445917594046118e-05,
+      "loss": 1.2233,
+      "step": 10698
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0381884740788584e-05,
+      "loss": 0.2396,
+      "step": 10699
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0318038003007969e-05,
+      "loss": 1.1151,
+      "step": 10700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0254377513298409e-05,
+      "loss": 0.1657,
+      "step": 10701
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.019090340386789e-05,
+      "loss": 0.7749,
+      "step": 10702
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0127615806536706e-05,
+      "loss": 2.1459,
+      "step": 10703
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0064514852738427e-05,
+      "loss": 2.4444,
+      "step": 10704
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.0001600673518408e-05,
+      "loss": 4.8659,
+      "step": 10705
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.93887339953471e-06,
+      "loss": 1.0776,
+      "step": 10706
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.876333161056617e-06,
+      "loss": 1.9856,
+      "step": 10707
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.813980087965635e-06,
+      "loss": 1.5399,
+      "step": 10708
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.751814309754015e-06,
+      "loss": 2.4091,
+      "step": 10709
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.689835955525317e-06,
+      "loss": 1.5442,
+      "step": 10710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.628045153993847e-06,
+      "loss": 4.3907,
+      "step": 10711
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.566442033484458e-06,
+      "loss": 0.6401,
+      "step": 10712
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.505026721931897e-06,
+      "loss": 3.9867,
+      "step": 10713
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.443799346881466e-06,
+      "loss": 1.5168,
+      "step": 10714
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.382760035487536e-06,
+      "loss": 0.3189,
+      "step": 10715
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.321908914514498e-06,
+      "loss": 0.7012,
+      "step": 10716
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.261246110335353e-06,
+      "loss": 0.7831,
+      "step": 10717
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.20077174893259e-06,
+      "loss": 0.6441,
+      "step": 10718
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.140485955896738e-06,
+      "loss": 1.7531,
+      "step": 10719
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.080388856427346e-06,
+      "loss": 0.3808,
+      "step": 10720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.02048057533147e-06,
+      "loss": 2.5578,
+      "step": 10721
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.960761237024629e-06,
+      "loss": 2.2826,
+      "step": 10722
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.901230965529361e-06,
+      "loss": 3.3241,
+      "step": 10723
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.841889884476173e-06,
+      "loss": 3.5025,
+      "step": 10724
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.782738117102051e-06,
+      "loss": 1.2396,
+      "step": 10725
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.723775786251409e-06,
+      "loss": 1.7128,
+      "step": 10726
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.665003014374761e-06,
+      "loss": 1.3706,
+      "step": 10727
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.606419923529207e-06,
+      "loss": 1.0153,
+      "step": 10728
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.548026635377892e-06,
+      "loss": 2.7404,
+      "step": 10729
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.48982327118968e-06,
+      "loss": 1.0336,
+      "step": 10730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.431809951839298e-06,
+      "loss": 1.9082,
+      "step": 10731
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.373986797806333e-06,
+      "loss": 0.1361,
+      "step": 10732
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.316353929175958e-06,
+      "loss": 0.758,
+      "step": 10733
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.258911465637642e-06,
+      "loss": 1.4646,
+      "step": 10734
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.201659526485982e-06,
+      "loss": 0.2779,
+      "step": 10735
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.144598230619337e-06,
+      "loss": 1.3526,
+      "step": 10736
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.087727696540693e-06,
+      "loss": 0.5334,
+      "step": 10737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.031048042356338e-06,
+      "loss": 2.7037,
+      "step": 10738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.974559385776681e-06,
+      "loss": 2.058,
+      "step": 10739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.918261844114993e-06,
+      "loss": 2.351,
+      "step": 10740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.86215553428794e-06,
+      "loss": 2.0215,
+      "step": 10741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.806240572814949e-06,
+      "loss": 4.8132,
+      "step": 10742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.750517075818154e-06,
+      "loss": 2.4494,
+      "step": 10743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.694985159021773e-06,
+      "loss": 1.5205,
+      "step": 10744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639644937752633e-06,
+      "loss": 0.8832,
+      "step": 10745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.584496526938934e-06,
+      "loss": 3.1433,
+      "step": 10746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.52954004111105e-06,
+      "loss": 1.9847,
+      "step": 10747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.474775594400241e-06,
+      "loss": 0.0941,
+      "step": 10748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.420203300539486e-06,
+      "loss": 0.0836,
+      "step": 10749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.3658232728621825e-06,
+      "loss": 2.7721,
+      "step": 10750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.311635624302937e-06,
+      "loss": 4.0772,
+      "step": 10751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.257640467396343e-06,
+      "loss": 2.943,
+      "step": 10752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.203837914277744e-06,
+      "loss": 0.9373,
+      "step": 10753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.1502280766819415e-06,
+      "loss": 3.7965,
+      "step": 10754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.096811065944076e-06,
+      "loss": 2.053,
+      "step": 10755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.043586992998286e-06,
+      "loss": 1.5233,
+      "step": 10756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.990555968378565e-06,
+      "loss": 1.9116,
+      "step": 10757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.937718102217506e-06,
+      "loss": 3.6914,
+      "step": 10758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.8850735042468415e-06,
+      "loss": 3.0371,
+      "step": 10759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.832622283796874e-06,
+      "loss": 0.1009,
+      "step": 10760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.780364549796337e-06,
+      "loss": 2.0524,
+      "step": 10761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.728300410771904e-06,
+      "loss": 2.1635,
+      "step": 10762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.676429974848541e-06,
+      "loss": 0.6747,
+      "step": 10763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.62475334974888e-06,
+      "loss": 1.1939,
+      "step": 10764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.573270642792739e-06,
+      "loss": 1.0864,
+      "step": 10765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.521981960897672e-06,
+      "loss": 0.7495,
+      "step": 10766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.470887410577809e-06,
+      "loss": 1.4194,
+      "step": 10767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.419987097944602e-06,
+      "loss": 6.2729,
+      "step": 10768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.369281128705629e-06,
+      "loss": 1.0879,
+      "step": 10769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.318769608165354e-06,
+      "loss": 0.4489,
+      "step": 10770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.2684526412239875e-06,
+      "loss": 0.9333,
+      "step": 10771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.218330332377964e-06,
+      "loss": 1.7493,
+      "step": 10772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.1684027857193516e-06,
+      "loss": 3.3385,
+      "step": 10773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.118670104935786e-06,
+      "loss": 0.0996,
+      "step": 10774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.0691323933101955e-06,
+      "loss": 1.6287,
+      "step": 10775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.0197897537206615e-06,
+      "loss": 4.5723,
+      "step": 10776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.970642288639916e-06,
+      "loss": 1.7212,
+      "step": 10777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.921690100135779e-06,
+      "loss": 5.3709,
+      "step": 10778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.872933289870042e-06,
+      "loss": 1.3991,
+      "step": 10779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.824371959099217e-06,
+      "loss": 1.3355,
+      "step": 10780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.776006208673401e-06,
+      "loss": 3.9192,
+      "step": 10781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.727836139036957e-06,
+      "loss": 2.2762,
+      "step": 10782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.679861850227408e-06,
+      "loss": 1.1704,
+      "step": 10783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.6320834418761796e-06,
+      "loss": 2.6363,
+      "step": 10784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.584501013207377e-06,
+      "loss": 0.184,
+      "step": 10785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.537114663038622e-06,
+      "loss": 1.6329,
+      "step": 10786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.489924489779852e-06,
+      "loss": 1.0647,
+      "step": 10787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.442930591434025e-06,
+      "loss": 1.1017,
+      "step": 10788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.3961330655961205e-06,
+      "loss": 0.4525,
+      "step": 10789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.349532009453484e-06,
+      "loss": 2.6384,
+      "step": 10790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.303127519785434e-06,
+      "loss": 1.3942,
+      "step": 10791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.256919692963036e-06,
+      "loss": 2.0443,
+      "step": 10792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.210908624948929e-06,
+      "loss": 1.7704,
+      "step": 10793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.165094411297089e-06,
+      "loss": 0.1986,
+      "step": 10794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.1194771471528844e-06,
+      "loss": 0.1671,
+      "step": 10795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.074056927252302e-06,
+      "loss": 2.4707,
+      "step": 10796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.02883384592252e-06,
+      "loss": 2.0439,
+      "step": 10797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.983807997080891e-06,
+      "loss": 1.1515,
+      "step": 10798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.938979474235617e-06,
+      "loss": 1.7813,
+      "step": 10799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.894348370484603e-06,
+      "loss": 2.0318,
+      "step": 10800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.849914778516273e-06,
+      "loss": 3.3639,
+      "step": 10801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.805678790608381e-06,
+      "loss": 3.6039,
+      "step": 10802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.76164049862875e-06,
+      "loss": 2.5012,
+      "step": 10803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.7177999940342554e-06,
+      "loss": 2.8527,
+      "step": 10804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.6741573678712565e-06,
+      "loss": 3.7209,
+      "step": 10805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.630712710775065e-06,
+      "loss": 2.5726,
+      "step": 10806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.5874661129699624e-06,
+      "loss": 1.5931,
+      "step": 10807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.544417664268652e-06,
+      "loss": 0.2384,
+      "step": 10808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.501567454072731e-06,
+      "loss": 2.2862,
+      "step": 10809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.458915571371647e-06,
+      "loss": 1.6151,
+      "step": 10810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.4164621047434465e-06,
+      "loss": 1.086,
+      "step": 10811
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.3742071423536145e-06,
+      "loss": 3.1173,
+      "step": 10812
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.3321507719558894e-06,
+      "loss": 0.1678,
+      "step": 10813
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2902930808910945e-06,
+      "loss": 1.8024,
+      "step": 10814
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.248634156087916e-06,
+      "loss": 1.9171,
+      "step": 10815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.207174084061793e-06,
+      "loss": 2.1006,
+      "step": 10816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.165912950915629e-06,
+      "loss": 0.966,
+      "step": 10817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.124850842338768e-06,
+      "loss": 5.4545,
+      "step": 10818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.083987843607618e-06,
+      "loss": 1.5345,
+      "step": 10819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0433240395847614e-06,
+      "loss": 2.2556,
+      "step": 10820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.0028595147192704e-06,
+      "loss": 0.4912,
+      "step": 10821
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.962594353046323e-06,
+      "loss": 1.6227,
+      "step": 10822
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.922528638187062e-06,
+      "loss": 0.2197,
+      "step": 10823
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.882662453348451e-06,
+      "loss": 0.6005,
+      "step": 10824
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.842995881323097e-06,
+      "loss": 0.1398,
+      "step": 10825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.8035290044888814e-06,
+      "loss": 2.6035,
+      "step": 10826
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.7642619048091854e-06,
+      "loss": 0.3045,
+      "step": 10827
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.725194663832543e-06,
+      "loss": 1.8866,
+      "step": 10828
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.686327362692088e-06,
+      "loss": 1.5495,
+      "step": 10829
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6476600821061634e-06,
+      "loss": 0.467,
+      "step": 10830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6091929023772896e-06,
+      "loss": 1.7644,
+      "step": 10831
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5709259033928298e-06,
+      "loss": 2.6947,
+      "step": 10832
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.532859164624036e-06,
+      "loss": 3.9461,
+      "step": 10833
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.494992765126659e-06,
+      "loss": 1.2752,
+      "step": 10834
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.457326783540027e-06,
+      "loss": 0.9796,
+      "step": 10835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.4198612980874677e-06,
+      "loss": 4.6836,
+      "step": 10836
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.382596386575876e-06,
+      "loss": 1.6318,
+      "step": 10837
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.345532126395634e-06,
+      "loss": 0.8507,
+      "step": 10838
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.3086685945202366e-06,
+      "loss": 1.929,
+      "step": 10839
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.2720058675066557e-06,
+      "loss": 1.2463,
+      "step": 10840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.235544021494452e-06,
+      "loss": 0.7559,
+      "step": 10841
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.1992831322063764e-06,
+      "loss": 1.3973,
+      "step": 10842
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.16322327494748e-06,
+      "loss": 1.9197,
+      "step": 10843
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.127364524605669e-06,
+      "loss": 3.0555,
+      "step": 10844
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.091706955650853e-06,
+      "loss": 2.9455,
+      "step": 10845
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.056250642135483e-06,
+      "loss": 1.1453,
+      "step": 10846
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0209956576937037e-06,
+      "loss": 2.3295,
+      "step": 10847
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.985942075541881e-06,
+      "loss": 0.7385,
+      "step": 10848
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.951089968477805e-06,
+      "loss": 3.2583,
+      "step": 10849
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.916439408881144e-06,
+      "loss": 1.4841,
+      "step": 10850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.881990468712703e-06,
+      "loss": 0.8721,
+      "step": 10851
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.847743219514898e-06,
+      "loss": 1.4982,
+      "step": 10852
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.813697732410969e-06,
+      "loss": 0.242,
+      "step": 10853
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.7798540781053263e-06,
+      "loss": 0.3177,
+      "step": 10854
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.7462123268831798e-06,
+      "loss": 2.7658,
+      "step": 10855
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.7127725486104647e-06,
+      "loss": 2.6335,
+      "step": 10856
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.6795348127335305e-06,
+      "loss": 3.6303,
+      "step": 10857
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.6464991882793167e-06,
+      "loss": 0.2245,
+      "step": 10858
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.6136657438550095e-06,
+      "loss": 2.7603,
+      "step": 10859
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.581034547647698e-06,
+      "loss": 1.5696,
+      "step": 10860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.5486056674247637e-06,
+      "loss": 1.5093,
+      "step": 10861
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.5163791705331117e-06,
+      "loss": 2.7892,
+      "step": 10862
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.484355123899651e-06,
+      "loss": 0.606,
+      "step": 10863
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.452533594030537e-06,
+      "loss": 1.0526,
+      "step": 10864
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.420914647011674e-06,
+      "loss": 1.8103,
+      "step": 10865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.389498348507968e-06,
+      "loss": 1.3051,
+      "step": 10866
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.3582847637636072e-06,
+      "loss": 1.9392,
+      "step": 10867
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.3272739576017945e-06,
+      "loss": 3.039,
+      "step": 10868
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.296465994424568e-06,
+      "loss": 2.4287,
+      "step": 10869
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.2658609382127384e-06,
+      "loss": 2.265,
+      "step": 10870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.2354588525257515e-06,
+      "loss": 1.9281,
+      "step": 10871
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.20525980050138e-06,
+      "loss": 1.5103,
+      "step": 10872
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.1752638448560126e-06,
+      "loss": 5.1444,
+      "step": 10873
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.1454710478839625e-06,
+      "loss": 2.2274,
+      "step": 10874
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.1158814714579146e-06,
+      "loss": 2.3204,
+      "step": 10875
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.086495177028225e-06,
+      "loss": 3.4075,
+      "step": 10876
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.0573122256233756e-06,
+      "loss": 2.7002,
+      "step": 10877
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.028332677849243e-06,
+      "loss": 3.1564,
+      "step": 10878
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9995565938896064e-06,
+      "loss": 2.0978,
+      "step": 10879
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9709840335053965e-06,
+      "loss": 2.0024,
+      "step": 10880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.942615056035135e-06,
+      "loss": 3.7339,
+      "step": 10881
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9144497203943067e-06,
+      "loss": 0.7228,
+      "step": 10882
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8864880850757437e-06,
+      "loss": 3.1147,
+      "step": 10883
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8587302081489954e-06,
+      "loss": 1.2827,
+      "step": 10884
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8311761472605936e-06,
+      "loss": 1.8742,
+      "step": 10885
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8038259596337537e-06,
+      "loss": 0.1293,
+      "step": 10886
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.77667970206834e-06,
+      "loss": 1.051,
+      "step": 10887
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.7497374309405568e-06,
+      "loss": 3.2292,
+      "step": 10888
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.7229992022032237e-06,
+      "loss": 1.507,
+      "step": 10889
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6964650713851338e-06,
+      "loss": 1.5165,
+      "step": 10890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6701350935914183e-06,
+      "loss": 3.3021,
+      "step": 10891
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6440093235032484e-06,
+      "loss": 1.0357,
+      "step": 10892
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6180878153774893e-06,
+      "loss": 3.2461,
+      "step": 10893
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.592370623047057e-06,
+      "loss": 0.3484,
+      "step": 10894
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.5668577999203181e-06,
+      "loss": 2.1864,
+      "step": 10895
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.5415493989814567e-06,
+      "loss": 2.5619,
+      "step": 10896
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.5164454727898847e-06,
+      "loss": 3.0296,
+      "step": 10897
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4915460734805097e-06,
+      "loss": 1.6482,
+      "step": 10898
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4668512527634681e-06,
+      "loss": 0.3049,
+      "step": 10899
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4423610619240025e-06,
+      "loss": 2.5762,
+      "step": 10900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4180755518224175e-06,
+      "loss": 0.7678,
+      "step": 10901
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.3939947728939584e-06,
+      "loss": 2.2315,
+      "step": 10902
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.3701187751485656e-06,
+      "loss": 0.351,
+      "step": 10903
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.346447608171142e-06,
+      "loss": 0.3683,
+      "step": 10904
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.3229813211209197e-06,
+      "loss": 2.6226,
+      "step": 10905
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2997199627319267e-06,
+      "loss": 3.2969,
+      "step": 10906
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2766635813123207e-06,
+      "loss": 1.9573,
+      "step": 10907
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2538122247448548e-06,
+      "loss": 1.7869,
+      "step": 10908
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.231165940486223e-06,
+      "loss": 2.1338,
+      "step": 10909
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.2087247755674824e-06,
+      "loss": 1.5464,
+      "step": 10910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.186488776593464e-06,
+      "loss": 3.0256,
+      "step": 10911
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.164457989743173e-06,
+      "loss": 2.7561,
+      "step": 10912
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.1426324607691664e-06,
+      "loss": 2.0451,
+      "step": 10913
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.121012234998009e-06,
+      "loss": 3.0774,
+      "step": 10914
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.0995973573296513e-06,
+      "loss": 2.5269,
+      "step": 10915
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.0783878722377405e-06,
+      "loss": 3.768,
+      "step": 10916
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.057383823769309e-06,
+      "loss": 1.4449,
+      "step": 10917
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.0365852555447753e-06,
+      "loss": 3.6626,
+      "step": 10918
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.015992210757799e-06,
+      "loss": 0.2236,
+      "step": 10919
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.956047321752592e-07,
+      "loss": 2.8395,
+      "step": 10920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.754228621370098e-07,
+      "loss": 0.5704,
+      "step": 10921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.55446642556057e-07,
+      "loss": 2.2406,
+      "step": 10922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.356761149182825e-07,
+      "loss": 1.2689,
+      "step": 10923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.161113202822646e-07,
+      "loss": 2.3906,
+      "step": 10924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.967522992795241e-07,
+      "loss": 0.2131,
+      "step": 10925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.775990921140453e-07,
+      "loss": 0.3812,
+      "step": 10926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.586517385625548e-07,
+      "loss": 1.8691,
+      "step": 10927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.399102779741097e-07,
+      "loss": 3.1741,
+      "step": 10928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.213747492703538e-07,
+      "loss": 4.9396,
+      "step": 10929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.030451909450842e-07,
+      "loss": 2.0551,
+      "step": 10930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.849216410644622e-07,
+      "loss": 2.0601,
+      "step": 10931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.670041372667802e-07,
+      "loss": 3.8174,
+      "step": 10932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.492927167624508e-07,
+      "loss": 1.1331,
+      "step": 10933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.317874163338068e-07,
+      "loss": 1.298,
+      "step": 10934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.144882723352564e-07,
+      "loss": 0.7721,
+      "step": 10935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.973953206929506e-07,
+      "loss": 3.5077,
+      "step": 10936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.805085969049274e-07,
+      "loss": 1.6516,
+      "step": 10937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.638281360408339e-07,
+      "loss": 5.8315,
+      "step": 10938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.473539727420708e-07,
+      "loss": 4.5834,
+      "step": 10939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.31086141221493e-07,
+      "loss": 1.6003,
+      "step": 10940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.150246752635758e-07,
+      "loss": 1.4007,
+      "step": 10941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.991696082240816e-07,
+      "loss": 2.1259,
+      "step": 10942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.835209730302938e-07,
+      "loss": 3.3716,
+      "step": 10943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.68078802180616e-07,
+      "loss": 0.1937,
+      "step": 10944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.528431277448288e-07,
+      "loss": 2.9024,
+      "step": 10945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.378139813637328e-07,
+      "loss": 3.2072,
+      "step": 10946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.229913942493613e-07,
+      "loss": 0.44,
+      "step": 10947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.08375397184635e-07,
+      "loss": 2.9909,
+      "step": 10948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.939660205235174e-07,
+      "loss": 2.0389,
+      "step": 10949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.797632941908492e-07,
+      "loss": 1.7857,
+      "step": 10950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.6576724768233606e-07,
+      "loss": 0.7787,
+      "step": 10951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.5197791006437174e-07,
+      "loss": 2.2154,
+      "step": 10952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.3839530997419333e-07,
+      "loss": 3.3343,
+      "step": 10953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.250194756195702e-07,
+      "loss": 0.8632,
+      "step": 10954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.118504347789376e-07,
+      "loss": 2.5128,
+      "step": 10955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.9888821480130733e-07,
+      "loss": 2.0711,
+      "step": 10956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.861328426060462e-07,
+      "loss": 2.005,
+      "step": 10957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7358434468308667e-07,
+      "loss": 0.0754,
+      "step": 10958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6124274709262717e-07,
+      "loss": 2.8036,
+      "step": 10959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4910807546525427e-07,
+      "loss": 1.7122,
+      "step": 10960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.3718035500177604e-07,
+      "loss": 0.2934,
+      "step": 10961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.2545961047322213e-07,
+      "loss": 1.9987,
+      "step": 10962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.139458662207995e-07,
+      "loss": 2.54,
+      "step": 10963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0263914615581424e-07,
+      "loss": 0.7984,
+      "step": 10964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.9153947375958333e-07,
+      "loss": 2.0604,
+      "step": 10965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8064687208356753e-07,
+      "loss": 2.262,
+      "step": 10966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.6996136374901613e-07,
+      "loss": 0.1892,
+      "step": 10967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.594829709472557e-07,
+      "loss": 1.7525,
+      "step": 10968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.492117154393681e-07,
+      "loss": 2.6991,
+      "step": 10969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.391476185563346e-07,
+      "loss": 1.5016,
+      "step": 10970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.2929070119884767e-07,
+      "loss": 2.373,
+      "step": 10971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.1964098383743248e-07,
+      "loss": 2.8329,
+      "step": 10972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.1019848651219199e-07,
+      "loss": 2.7789,
+      "step": 10973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0096322883299544e-07,
+      "loss": 1.7445,
+      "step": 10974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.919352299792232e-07,
+      "loss": 0.114,
+      "step": 10975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8311450869994416e-07,
+      "loss": 2.7529,
+      "step": 10976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7450108331361624e-07,
+      "loss": 2.0277,
+      "step": 10977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6609497170835264e-07,
+      "loss": 2.2584,
+      "step": 10978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5789619134157774e-07,
+      "loss": 3.2083,
+      "step": 10979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4990475924023807e-07,
+      "loss": 1.3959,
+      "step": 10980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4212069200063573e-07,
+      "loss": 1.0096,
+      "step": 10981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.345440057884173e-07,
+      "loss": 1.8895,
+      "step": 10982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2717471633854062e-07,
+      "loss": 2.3588,
+      "step": 10983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.200128389552968e-07,
+      "loss": 1.6414,
+      "step": 10984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.13058388512155e-07,
+      "loss": 3.7154,
+      "step": 10985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.063113794518844e-07,
+      "loss": 0.7348,
+      "step": 10986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.977182578643219e-08,
+      "loss": 3.9676,
+      "step": 10987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.343974109684572e-08,
+      "loss": 0.9204,
+      "step": 10988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.731513853340589e-08,
+      "loss": 1.1852,
+      "step": 10989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.139803081540498e-08,
+      "loss": 1.0281,
+      "step": 10990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.568843023130213e-08,
+      "loss": 0.6824,
+      "step": 10991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.018634863856788e-08,
+      "loss": 3.2435,
+      "step": 10992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.489179746370644e-08,
+      "loss": 1.5084,
+      "step": 10993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.980478770224452e-08,
+      "loss": 3.2341,
+      "step": 10994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.492532991868693e-08,
+      "loss": 1.701,
+      "step": 10995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.025343424651663e-08,
+      "loss": 0.5415,
+      "step": 10996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.578911038813916e-08,
+      "loss": 0.3403,
+      "step": 10997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.153236761488266e-08,
+      "loss": 0.7287,
+      "step": 10998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7483214767009e-08,
+      "loss": 3.3429,
+      "step": 10999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.364166025362492e-08,
+      "loss": 1.8966,
+      "step": 11000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0007712052737556e-08,
+      "loss": 1.2487,
+      "step": 11001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.6581377711176747e-08,
+      "loss": 0.9395,
+      "step": 11002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.3362664344639407e-08,
+      "loss": 0.3025,
+      "step": 11003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0351578637611833e-08,
+      "loss": 2.7826,
+      "step": 11004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7548126843425215e-08,
+      "loss": 1.2448,
+      "step": 11005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4952314784144607e-08,
+      "loss": 2.4257,
+      "step": 11006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2564147850691044e-08,
+      "loss": 0.7643,
+      "step": 11007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.0383631002686134e-08,
+      "loss": 3.2187,
+      "step": 11008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.41076876856306e-09,
+      "loss": 0.8638,
+      "step": 11009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.64556524547777e-09,
+      "loss": 0.4226,
+      "step": 11010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.088024099353383e-09,
+      "loss": 0.6444,
+      "step": 11011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.738148564813582e-09,
+      "loss": 2.1898,
+      "step": 11012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.595941445238115e-09,
+      "loss": 1.4152,
+      "step": 11013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6614051127072928e-09,
+      "loss": 1.0957,
+      "step": 11014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.345415080463936e-10,
+      "loss": 1.2377,
+      "step": 11015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1535214075905106e-10,
+      "loss": 3.5507,
+      "step": 11016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.038380891049684e-10,
+      "loss": 0.1457,
+      "step": 11017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0,
+      "loss": 3.0085,
+      "step": 11018
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.038380891049684e-10,
+      "loss": 2.3061,
+      "step": 11019
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1535214075905106e-10,
+      "loss": 2.2387,
+      "step": 11020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.345415080463936e-10,
+      "loss": 0.312,
+      "step": 11021
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6614051127072928e-09,
+      "loss": 1.7864,
+      "step": 11022
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.5959414452270124e-09,
+      "loss": 1.9194,
+      "step": 11023
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.738148564813582e-09,
+      "loss": 1.5768,
+      "step": 11024
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.088024099353383e-09,
+      "loss": 2.1877,
+      "step": 11025
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.64556524547777e-09,
+      "loss": 3.7383,
+      "step": 11026
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.41076876856306e-09,
+      "loss": 3.3486,
+      "step": 11027
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.0383631002686134e-08,
+      "loss": 2.3161,
+      "step": 11028
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.2564147850679942e-08,
+      "loss": 1.248,
+      "step": 11029
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4952314784144607e-08,
+      "loss": 1.8727,
+      "step": 11030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7548126843414115e-08,
+      "loss": 1.8818,
+      "step": 11031
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0351578637611833e-08,
+      "loss": 1.013,
+      "step": 11032
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.3362664344639407e-08,
+      "loss": 3.6049,
+      "step": 11033
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.6581377711176747e-08,
+      "loss": 0.1756,
+      "step": 11034
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0007712052737556e-08,
+      "loss": 2.6143,
+      "step": 11035
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.364166025362492e-08,
+      "loss": 1.158,
+      "step": 11036
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7483214767009e-08,
+      "loss": 1.1603,
+      "step": 11037
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.153236761488266e-08,
+      "loss": 1.0144,
+      "step": 11038
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.578911038813916e-08,
+      "loss": 2.9042,
+      "step": 11039
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.0253434246505526e-08,
+      "loss": 2.2586,
+      "step": 11040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.492532991868693e-08,
+      "loss": 0.8926,
+      "step": 11041
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.980478770223341e-08,
+      "loss": 4.0642,
+      "step": 11042
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.489179746369534e-08,
+      "loss": 1.1372,
+      "step": 11043
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.018634863855677e-08,
+      "loss": 2.8981,
+      "step": 11044
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.568843023130213e-08,
+      "loss": 1.7,
+      "step": 11045
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.139803081539387e-08,
+      "loss": 3.4021,
+      "step": 11046
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8.731513853339479e-08,
+      "loss": 1.9016,
+      "step": 11047
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.343974109684572e-08,
+      "loss": 1.6599,
+      "step": 11048
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.977182578642108e-08,
+      "loss": 0.8918,
+      "step": 11049
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.063113794518844e-07,
+      "loss": 1.985,
+      "step": 11050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.13058388512155e-07,
+      "loss": 3.5248,
+      "step": 11051
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.200128389552857e-07,
+      "loss": 2.586,
+      "step": 11052
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.271747163385295e-07,
+      "loss": 1.6413,
+      "step": 11053
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.345440057884062e-07,
+      "loss": 2.564,
+      "step": 11054
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.421206920006246e-07,
+      "loss": 1.1656,
+      "step": 11055
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4990475924023807e-07,
+      "loss": 2.0379,
+      "step": 11056
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5789619134157774e-07,
+      "loss": 2.8709,
+      "step": 11057
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6609497170834155e-07,
+      "loss": 2.5539,
+      "step": 11058
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7450108331361624e-07,
+      "loss": 0.8268,
+      "step": 11059
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8311450869993307e-07,
+      "loss": 2.0412,
+      "step": 11060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.919352299792232e-07,
+      "loss": 2.4593,
+      "step": 11061
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.0096322883298435e-07,
+      "loss": 3.1579,
+      "step": 11062
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.1019848651218087e-07,
+      "loss": 2.1402,
+      "step": 11063
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.1964098383742137e-07,
+      "loss": 3.2981,
+      "step": 11064
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.2929070119883655e-07,
+      "loss": 1.0304,
+      "step": 11065
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.391476185563235e-07,
+      "loss": 2.3136,
+      "step": 11066
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.4921171543935696e-07,
+      "loss": 0.9773,
+      "step": 11067
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.594829709472557e-07,
+      "loss": 1.4968,
+      "step": 11068
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.6996136374900507e-07,
+      "loss": 0.2808,
+      "step": 11069
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8064687208355646e-07,
+      "loss": 1.6033,
+      "step": 11070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.9153947375958333e-07,
+      "loss": 1.5,
+      "step": 11071
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.026391461558031e-07,
+      "loss": 0.6861,
+      "step": 11072
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.1394586622078836e-07,
+      "loss": 2.5708,
+      "step": 11073
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.2545961047321107e-07,
+      "loss": 1.4824,
+      "step": 11074
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.371803550017649e-07,
+      "loss": 2.7127,
+      "step": 11075
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4910807546525427e-07,
+      "loss": 3.7346,
+      "step": 11076
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.6124274709261605e-07,
+      "loss": 4.7117,
+      "step": 11077
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.7358434468307555e-07,
+      "loss": 1.5882,
+      "step": 11078
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.861328426060351e-07,
+      "loss": 0.456,
+      "step": 11079
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.988882148012962e-07,
+      "loss": 4.1593,
+      "step": 11080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.118504347789265e-07,
+      "loss": 2.8858,
+      "step": 11081
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.250194756195591e-07,
+      "loss": 1.2035,
+      "step": 11082
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.383953099741822e-07,
+      "loss": 0.785,
+      "step": 11083
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.519779100643606e-07,
+      "loss": 2.4256,
+      "step": 11084
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.6576724768232494e-07,
+      "loss": 0.6105,
+      "step": 11085
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.797632941908381e-07,
+      "loss": 2.1764,
+      "step": 11086
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.939660205235064e-07,
+      "loss": 2.1097,
+      "step": 11087
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.083753971846239e-07,
+      "loss": 2.9264,
+      "step": 11088
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.229913942493503e-07,
+      "loss": 0.5487,
+      "step": 11089
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.378139813637217e-07,
+      "loss": 0.9908,
+      "step": 11090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.528431277448177e-07,
+      "loss": 4.7229,
+      "step": 11091
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.68078802180605e-07,
+      "loss": 1.3365,
+      "step": 11092
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.835209730302827e-07,
+      "loss": 2.6963,
+      "step": 11093
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 5.991696082240705e-07,
+      "loss": 2.8086,
+      "step": 11094
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.150246752635647e-07,
+      "loss": 2.505,
+      "step": 11095
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.310861412214708e-07,
+      "loss": 3.6147,
+      "step": 11096
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.473539727420597e-07,
+      "loss": 3.1786,
+      "step": 11097
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.638281360408116e-07,
+      "loss": 0.7386,
+      "step": 11098
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.805085969049164e-07,
+      "loss": 3.1802,
+      "step": 11099
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.973953206929285e-07,
+      "loss": 1.1276,
+      "step": 11100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.144882723352453e-07,
+      "loss": 1.6761,
+      "step": 11101
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.317874163337846e-07,
+      "loss": 1.4855,
+      "step": 11102
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.492927167624398e-07,
+      "loss": 0.2296,
+      "step": 11103
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.670041372667691e-07,
+      "loss": 2.8755,
+      "step": 11104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.849216410644511e-07,
+      "loss": 2.8311,
+      "step": 11105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 8.030451909450731e-07,
+      "loss": 0.6817,
+      "step": 11106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 8.213747492703316e-07,
+      "loss": 0.7883,
+      "step": 11107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 8.399102779740986e-07,
+      "loss": 1.6138,
+      "step": 11108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 8.586517385625326e-07,
+      "loss": 1.6454,
+      "step": 11109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 8.775990921140231e-07,
+      "loss": 2.3825,
+      "step": 11110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 8.967522992795129e-07,
+      "loss": 2.5501,
+      "step": 11111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.161113202822536e-07,
+      "loss": 1.9934,
+      "step": 11112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.356761149182602e-07,
+      "loss": 2.2191,
+      "step": 11113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.55446642556035e-07,
+      "loss": 3.2105,
+      "step": 11114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.754228621369988e-07,
+      "loss": 1.4156,
+      "step": 11115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.95604732175237e-07,
+      "loss": 2.169,
+      "step": 11116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.0159922107577769e-06,
+      "loss": 0.8527,
+      "step": 11117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.036585255544753e-06,
+      "loss": 3.9015,
+      "step": 11118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.0573838237692979e-06,
+      "loss": 4.8941,
+      "step": 11119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.0783878722377183e-06,
+      "loss": 0.9793,
+      "step": 11120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.0995973573296403e-06,
+      "loss": 1.6421,
+      "step": 11121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1210122349979868e-06,
+      "loss": 0.8842,
+      "step": 11122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1426324607691551e-06,
+      "loss": 4.8496,
+      "step": 11123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.1644579897431507e-06,
+      "loss": 1.4914,
+      "step": 11124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.186488776593453e-06,
+      "loss": 1.1431,
+      "step": 11125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.2087247755674601e-06,
+      "loss": 0.8991,
+      "step": 11126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.2311659404862008e-06,
+      "loss": 0.7732,
+      "step": 11127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.2538122247448325e-06,
+      "loss": 3.9526,
+      "step": 11128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.2766635813123096e-06,
+      "loss": 3.7085,
+      "step": 11129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.2997199627319045e-06,
+      "loss": 1.9523,
+      "step": 11130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.3229813211209086e-06,
+      "loss": 2.9123,
+      "step": 11131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.3464476081711198e-06,
+      "loss": 1.1808,
+      "step": 11132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.3701187751485544e-06,
+      "loss": 2.3146,
+      "step": 11133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.3939947728939361e-06,
+      "loss": 1.2694,
+      "step": 11134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4180755518223955e-06,
+      "loss": 2.1986,
+      "step": 11135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4423610619239802e-06,
+      "loss": 1.921,
+      "step": 11136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.466851252763446e-06,
+      "loss": 0.0975,
+      "step": 11137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4915460734804877e-06,
+      "loss": 0.5928,
+      "step": 11138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5164454727898624e-06,
+      "loss": 3.8411,
+      "step": 11139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5415493989814344e-06,
+      "loss": 1.6572,
+      "step": 11140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5668577999202961e-06,
+      "loss": 0.9392,
+      "step": 11141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5923706230470347e-06,
+      "loss": 0.1446,
+      "step": 11142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.618087815377467e-06,
+      "loss": 0.6636,
+      "step": 11143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6440093235032262e-06,
+      "loss": 0.5087,
+      "step": 11144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6701350935913961e-06,
+      "loss": 1.804,
+      "step": 11145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.6964650713851116e-06,
+      "loss": 0.7728,
+      "step": 11146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7229992022032016e-06,
+      "loss": 1.2706,
+      "step": 11147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7497374309405346e-06,
+      "loss": 2.1872,
+      "step": 11148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.776679702068318e-06,
+      "loss": 0.0751,
+      "step": 11149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8038259596337314e-06,
+      "loss": 0.7912,
+      "step": 11150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8311761472605716e-06,
+      "loss": 1.1876,
+      "step": 11151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8587302081489732e-06,
+      "loss": 2.7664,
+      "step": 11152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8864880850757217e-06,
+      "loss": 2.179,
+      "step": 11153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9144497203942847e-06,
+      "loss": 1.4658,
+      "step": 11154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.942615056035113e-06,
+      "loss": 0.115,
+      "step": 11155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.970984033505363e-06,
+      "loss": 1.273,
+      "step": 11156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.9995565938895844e-06,
+      "loss": 0.6908,
+      "step": 11157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.028332677849221e-06,
+      "loss": 1.22,
+      "step": 11158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.0573122256233425e-06,
+      "loss": 2.901,
+      "step": 11159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.086495177028203e-06,
+      "loss": 2.2785,
+      "step": 11160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.115881471457892e-06,
+      "loss": 1.7848,
+      "step": 11161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.14547104788394e-06,
+      "loss": 1.3688,
+      "step": 11162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.17526384485599e-06,
+      "loss": 0.5316,
+      "step": 11163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.205259800501358e-06,
+      "loss": 3.9401,
+      "step": 11164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.235458852525718e-06,
+      "loss": 1.8591,
+      "step": 11165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.265860938212705e-06,
+      "loss": 2.3142,
+      "step": 11166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.296465994424546e-06,
+      "loss": 2.231,
+      "step": 11167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.327273957601761e-06,
+      "loss": 1.5636,
+      "step": 11168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.358284763763585e-06,
+      "loss": 2.5623,
+      "step": 11169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.389498348507935e-06,
+      "loss": 3.4586,
+      "step": 11170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.4209146470116516e-06,
+      "loss": 0.2209,
+      "step": 11171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.452533594030515e-06,
+      "loss": 2.779,
+      "step": 11172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.484355123899618e-06,
+      "loss": 4.0235,
+      "step": 11173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.5163791705330787e-06,
+      "loss": 2.1683,
+      "step": 11174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.5486056674247307e-06,
+      "loss": 2.305,
+      "step": 11175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.581034547647665e-06,
+      "loss": 0.1125,
+      "step": 11176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.613665743854976e-06,
+      "loss": 0.9971,
+      "step": 11177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.6464991882792833e-06,
+      "loss": 1.0252,
+      "step": 11178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.6795348127335085e-06,
+      "loss": 2.3466,
+      "step": 11179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.7127725486104427e-06,
+      "loss": 1.168,
+      "step": 11180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.7462123268831463e-06,
+      "loss": 1.6468,
+      "step": 11181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.779854078105293e-06,
+      "loss": 3.3546,
+      "step": 11182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.813697732410936e-06,
+      "loss": 2.9254,
+      "step": 11183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8477432195148644e-06,
+      "loss": 1.7207,
+      "step": 11184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8819904687126806e-06,
+      "loss": 1.08,
+      "step": 11185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9164394088811107e-06,
+      "loss": 2.103,
+      "step": 11186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9510899684777715e-06,
+      "loss": 2.6141,
+      "step": 11187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9859420755418588e-06,
+      "loss": 1.8891,
+      "step": 11188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0209956576936706e-06,
+      "loss": 2.3763,
+      "step": 11189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0562506421354496e-06,
+      "loss": 0.1122,
+      "step": 11190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0917069556508194e-06,
+      "loss": 1.7106,
+      "step": 11191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.127364524605636e-06,
+      "loss": 0.4592,
+      "step": 11192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.1632232749474466e-06,
+      "loss": 3.117,
+      "step": 11193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.1992831322063434e-06,
+      "loss": 2.5648,
+      "step": 11194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2355440214944298e-06,
+      "loss": 3.3029,
+      "step": 11195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2720058675066222e-06,
+      "loss": 2.2226,
+      "step": 11196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3086685945202036e-06,
+      "loss": 2.0321,
+      "step": 11197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3455321263956007e-06,
+      "loss": 0.5903,
+      "step": 11198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3825963865758425e-06,
+      "loss": 2.3343,
+      "step": 11199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4198612980874346e-06,
+      "loss": 1.1682,
+      "step": 11200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.4573267835399934e-06,
+      "loss": 0.9921,
+      "step": 11201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.494992765126626e-06,
+      "loss": 0.9815,
+      "step": 11202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5328591646240026e-06,
+      "loss": 4.0717,
+      "step": 11203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.5709259033927967e-06,
+      "loss": 1.4389,
+      "step": 11204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.609192902377256e-06,
+      "loss": 0.1594,
+      "step": 11205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.6476600821061303e-06,
+      "loss": 0.6133,
+      "step": 11206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.686327362692055e-06,
+      "loss": 4.0144,
+      "step": 11207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.72519466383251e-06,
+      "loss": 1.3137,
+      "step": 11208
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.764261904809152e-06,
+      "loss": 1.0408,
+      "step": 11209
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.8035290044888484e-06,
+      "loss": 1.5411,
+      "step": 11210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.8429958813230635e-06,
+      "loss": 0.9301,
+      "step": 11211
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.882662453348418e-06,
+      "loss": 1.9214,
+      "step": 11212
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.922528638187028e-06,
+      "loss": 2.0311,
+      "step": 11213
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.962594353046279e-06,
+      "loss": 2.123,
+      "step": 11214
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.002859514719226e-06,
+      "loss": 3.3439,
+      "step": 11215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.043324039584728e-06,
+      "loss": 1.8588,
+      "step": 11216
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.083987843607584e-06,
+      "loss": 0.3653,
+      "step": 11217
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.124850842338735e-06,
+      "loss": 2.2308,
+      "step": 11218
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.165912950915596e-06,
+      "loss": 2.0534,
+      "step": 11219
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.20717408406176e-06,
+      "loss": 3.4084,
+      "step": 11220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.248634156087927e-06,
+      "loss": 0.9603,
+      "step": 11221
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.2902930808911165e-06,
+      "loss": 3.1939,
+      "step": 11222
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.3321507719558005e-06,
+      "loss": 1.7248,
+      "step": 11223
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.3742071423536365e-06,
+      "loss": 0.5832,
+      "step": 11224
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.4164621047433575e-06,
+      "loss": 0.1347,
+      "step": 11225
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.458915571371558e-06,
+      "loss": 0.1231,
+      "step": 11226
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.501567454072642e-06,
+      "loss": 3.8183,
+      "step": 11227
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.544417664268563e-06,
+      "loss": 2.0563,
+      "step": 11228
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.5874661129698735e-06,
+      "loss": 3.0561,
+      "step": 11229
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.630712710774976e-06,
+      "loss": 0.1707,
+      "step": 11230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.6741573678712125e-06,
+      "loss": 0.4315,
+      "step": 11231
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.7177999940341665e-06,
+      "loss": 2.0292,
+      "step": 11232
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.7616404986287056e-06,
+      "loss": 1.6124,
+      "step": 11233
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.8056787906083364e-06,
+      "loss": 3.0637,
+      "step": 11234
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.84991477851624e-06,
+      "loss": 0.1562,
+      "step": 11235
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.8943483704845696e-06,
+      "loss": 1.9127,
+      "step": 11236
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.938979474235572e-06,
+      "loss": 2.2293,
+      "step": 11237
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.983807997080858e-06,
+      "loss": 1.871,
+      "step": 11238
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.028833845922476e-06,
+      "loss": 0.0967,
+      "step": 11239
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.074056927252258e-06,
+      "loss": 2.3518,
+      "step": 11240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.119477147152851e-06,
+      "loss": 2.2615,
+      "step": 11241
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.165094411297044e-06,
+      "loss": 2.1384,
+      "step": 11242
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.210908624948885e-06,
+      "loss": 0.3412,
+      "step": 11243
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.256919692963047e-06,
+      "loss": 1.5205,
+      "step": 11244
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.3031275197854e-06,
+      "loss": 1.7618,
+      "step": 11245
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.349532009453506e-06,
+      "loss": 2.2433,
+      "step": 11246
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.3961330655960764e-06,
+      "loss": 4.7605,
+      "step": 11247
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.442930591434048e-06,
+      "loss": 0.6156,
+      "step": 11248
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.489924489779863e-06,
+      "loss": 2.0257,
+      "step": 11249
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.537114663038645e-06,
+      "loss": 1.7652,
+      "step": 11250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.584501013207399e-06,
+      "loss": 1.0858,
+      "step": 11251
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.632083441876191e-06,
+      "loss": 0.7042,
+      "step": 11252
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.67986185022743e-06,
+      "loss": 2.4927,
+      "step": 11253
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.727836139036857e-06,
+      "loss": 0.3653,
+      "step": 11254
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.776006208673423e-06,
+      "loss": 2.2077,
+      "step": 11255
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.824371959099117e-06,
+      "loss": 1.6429,
+      "step": 11256
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.872933289869942e-06,
+      "loss": 3.1049,
+      "step": 11257
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.921690100135679e-06,
+      "loss": 2.5328,
+      "step": 11258
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.970642288639816e-06,
+      "loss": 2.4032,
+      "step": 11259
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.0197897537205615e-06,
+      "loss": 0.1692,
+      "step": 11260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.0691323933100845e-06,
+      "loss": 0.6793,
+      "step": 11261
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.118670104935742e-06,
+      "loss": 0.8851,
+      "step": 11262
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.168402785719241e-06,
+      "loss": 0.5812,
+      "step": 11263
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.21833033237792e-06,
+      "loss": 1.0591,
+      "step": 11264
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.268452641223888e-06,
+      "loss": 2.6147,
+      "step": 11265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.318769608165309e-06,
+      "loss": 0.229,
+      "step": 11266
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.369281128705584e-06,
+      "loss": 3.208,
+      "step": 11267
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.419987097944557e-06,
+      "loss": 0.3165,
+      "step": 11268
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.470887410577764e-06,
+      "loss": 2.7876,
+      "step": 11269
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.521981960897627e-06,
+      "loss": 2.4971,
+      "step": 11270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.573270642792695e-06,
+      "loss": 1.2758,
+      "step": 11271
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.624753349748847e-06,
+      "loss": 2.3055,
+      "step": 11272
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.676429974848497e-06,
+      "loss": 4.6844,
+      "step": 11273
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.72830041077186e-06,
+      "loss": 2.1461,
+      "step": 11274
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.780364549796292e-06,
+      "loss": 1.6397,
+      "step": 11275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83262228379683e-06,
+      "loss": 1.4292,
+      "step": 11276
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.885073504246864e-06,
+      "loss": 3.5938,
+      "step": 11277
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.937718102217461e-06,
+      "loss": 0.7048,
+      "step": 11278
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.990555968378587e-06,
+      "loss": 3.1485,
+      "step": 11279
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.043586992998308e-06,
+      "loss": 1.0309,
+      "step": 11280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096811065944098e-06,
+      "loss": 2.6656,
+      "step": 11281
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.150228076681964e-06,
+      "loss": 2.407,
+      "step": 11282
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.203837914277755e-06,
+      "loss": 1.2758,
+      "step": 11283
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.257640467396365e-06,
+      "loss": 2.6698,
+      "step": 11284
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.311635624302826e-06,
+      "loss": 1.9189,
+      "step": 11285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.3658232728622046e-06,
+      "loss": 1.1321,
+      "step": 11286
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.420203300539375e-06,
+      "loss": 1.4722,
+      "step": 11287
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.474775594400129e-06,
+      "loss": 3.5565,
+      "step": 11288
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.529540041110939e-06,
+      "loss": 1.9717,
+      "step": 11289
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.584496526938811e-06,
+      "loss": 1.8248,
+      "step": 11290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.639644937752521e-06,
+      "loss": 2.0974,
+      "step": 11291
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.694985159021651e-06,
+      "loss": 1.2829,
+      "step": 11292
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.75051707581803e-06,
+      "loss": 0.7763,
+      "step": 11293
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.806240572814837e-06,
+      "loss": 0.2481,
+      "step": 11294
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.862155534287896e-06,
+      "loss": 1.6131,
+      "step": 11295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.918261844114883e-06,
+      "loss": 0.7102,
+      "step": 11296
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.974559385776637e-06,
+      "loss": 1.0198,
+      "step": 11297
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.031048042356294e-06,
+      "loss": 0.9501,
+      "step": 11298
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.087727696540649e-06,
+      "loss": 2.1273,
+      "step": 11299
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.144598230619283e-06,
+      "loss": 4.3872,
+      "step": 11300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.201659526485938e-06,
+      "loss": 1.4658,
+      "step": 11301
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.258911465637597e-06,
+      "loss": 2.3358,
+      "step": 11302
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.316353929175912e-06,
+      "loss": 1.2043,
+      "step": 11303
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.373986797806277e-06,
+      "loss": 2.6059,
+      "step": 11304
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.431809951839253e-06,
+      "loss": 4.5548,
+      "step": 11305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.489823271189623e-06,
+      "loss": 2.6919,
+      "step": 11306
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.548026635377836e-06,
+      "loss": 3.9822,
+      "step": 11307
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.60641992352923e-06,
+      "loss": 0.8613,
+      "step": 11308
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.665003014374706e-06,
+      "loss": 3.5767,
+      "step": 11309
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.723775786251431e-06,
+      "loss": 2.1878,
+      "step": 11310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.782738117102073e-06,
+      "loss": 0.1258,
+      "step": 11311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.841889884476195e-06,
+      "loss": 1.0038,
+      "step": 11312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.901230965529383e-06,
+      "loss": 1.5331,
+      "step": 11313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 8.960761237024651e-06,
+      "loss": 1.646,
+      "step": 11314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.020480575331491e-06,
+      "loss": 1.0134,
+      "step": 11315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.080388856427225e-06,
+      "loss": 3.5519,
+      "step": 11316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.14048595589676e-06,
+      "loss": 1.1977,
+      "step": 11317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.200771748932457e-06,
+      "loss": 0.8974,
+      "step": 11318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.261246110335375e-06,
+      "loss": 2.3879,
+      "step": 11319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.321908914514376e-06,
+      "loss": 0.9738,
+      "step": 11320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.382760035487404e-06,
+      "loss": 0.1256,
+      "step": 11321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.443799346881344e-06,
+      "loss": 1.0877,
+      "step": 11322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.505026721931776e-06,
+      "loss": 2.7353,
+      "step": 11323
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.566442033484324e-06,
+      "loss": 0.525,
+      "step": 11324
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.628045153993725e-06,
+      "loss": 0.603,
+      "step": 11325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.689835955525273e-06,
+      "loss": 0.635,
+      "step": 11326
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.751814309753881e-06,
+      "loss": 1.5161,
+      "step": 11327
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.813980087965591e-06,
+      "loss": 1.9773,
+      "step": 11328
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.876333161056572e-06,
+      "loss": 2.2358,
+      "step": 11329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.938873399534654e-06,
+      "loss": 2.751,
+      "step": 11330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0001600673518352e-05,
+      "loss": 0.5908,
+      "step": 11331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0064514852738383e-05,
+      "loss": 2.4435,
+      "step": 11332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.012761580653665e-05,
+      "loss": 1.0205,
+      "step": 11333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0190903403867836e-05,
+      "loss": 1.0669,
+      "step": 11334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0254377513298353e-05,
+      "loss": 4.7681,
+      "step": 11335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0318038003007913e-05,
+      "loss": 0.9447,
+      "step": 11336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0381884740788538e-05,
+      "loss": 2.364,
+      "step": 11337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0445917594046062e-05,
+      "loss": 1.707,
+      "step": 11338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0510136429799355e-05,
+      "loss": 3.2172,
+      "step": 11339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0574541114681037e-05,
+      "loss": 0.5928,
+      "step": 11340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0639131514938228e-05,
+      "loss": 4.2026,
+      "step": 11341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0703907496431741e-05,
+      "loss": 0.2068,
+      "step": 11342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.076886892463762e-05,
+      "loss": 1.0083,
+      "step": 11343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0834015664646124e-05,
+      "loss": 1.5276,
+      "step": 11344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0899347581163311e-05,
+      "loss": 2.1582,
+      "step": 11345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0964864538510033e-05,
+      "loss": 2.9985,
+      "step": 11346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.103056640062351e-05,
+      "loss": 1.4129,
+      "step": 11347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1096453031056286e-05,
+      "loss": 3.1987,
+      "step": 11348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1162524292977705e-05,
+      "loss": 2.9337,
+      "step": 11349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1228780049173637e-05,
+      "loss": 2.9989,
+      "step": 11350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.129522016204636e-05,
+      "loss": 1.0688,
+      "step": 11351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1361844493615725e-05,
+      "loss": 2.5366,
+      "step": 11352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1428652905519044e-05,
+      "loss": 0.1448,
+      "step": 11353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1495645259010834e-05,
+      "loss": 0.6196,
+      "step": 11354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1562821414964253e-05,
+      "loss": 1.0933,
+      "step": 11355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1630181233870107e-05,
+      "loss": 0.9082,
+      "step": 11356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1697724575838431e-05,
+      "loss": 2.3748,
+      "step": 11357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1765451300597452e-05,
+      "loss": 1.4672,
+      "step": 11358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1833361267495235e-05,
+      "loss": 2.9499,
+      "step": 11359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.190145433549863e-05,
+      "loss": 1.4595,
+      "step": 11360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.1969730363194875e-05,
+      "loss": 0.9311,
+      "step": 11361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2038189208790607e-05,
+      "loss": 1.3365,
+      "step": 11362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2106830730113428e-05,
+      "loss": 4.6457,
+      "step": 11363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2175654784610924e-05,
+      "loss": 2.2725,
+      "step": 11364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2244661229352271e-05,
+      "loss": 0.2012,
+      "step": 11365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2313849921027177e-05,
+      "loss": 0.9252,
+      "step": 11366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2383220715947551e-05,
+      "loss": 0.5369,
+      "step": 11367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2452773470046441e-05,
+      "loss": 0.1765,
+      "step": 11368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2522508038879687e-05,
+      "loss": 1.6837,
+      "step": 11369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2592424277624848e-05,
+      "loss": 4.2423,
+      "step": 11370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2662522041082914e-05,
+      "loss": 1.1769,
+      "step": 11371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2732801183677335e-05,
+      "loss": 2.7184,
+      "step": 11372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2803261559454916e-05,
+      "loss": 0.8993,
+      "step": 11373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.2873903022086586e-05,
+      "loss": 0.9965,
+      "step": 11374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.29447254248665e-05,
+      "loss": 1.2813,
+      "step": 11375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3015728620713751e-05,
+      "loss": 1.8898,
+      "step": 11376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3086912462171274e-05,
+      "loss": 0.3567,
+      "step": 11377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3158276801407521e-05,
+      "loss": 1.3808,
+      "step": 11378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3229821490215433e-05,
+      "loss": 2.4499,
+      "step": 11379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3301546380013885e-05,
+      "loss": 4.3667,
+      "step": 11380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3373451321847552e-05,
+      "loss": 0.8275,
+      "step": 11381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3445536166386651e-05,
+      "loss": 2.6284,
+      "step": 11382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3517800763928312e-05,
+      "loss": 1.5506,
+      "step": 11383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3590244964396348e-05,
+      "loss": 2.3424,
+      "step": 11384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3662868617341051e-05,
+      "loss": 1.6745,
+      "step": 11385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3735671571940766e-05,
+      "loss": 4.1679,
+      "step": 11386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.380865367700075e-05,
+      "loss": 0.3612,
+      "step": 11387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.3881814780954882e-05,
+      "loss": 3.7309,
+      "step": 11388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.395515473186464e-05,
+      "loss": 2.6183,
+      "step": 11389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4028673377420776e-05,
+      "loss": 1.494,
+      "step": 11390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4102370564942213e-05,
+      "loss": 1.3039,
+      "step": 11391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4176246141377814e-05,
+      "loss": 3.9451,
+      "step": 11392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4250299953305235e-05,
+      "loss": 0.8824,
+      "step": 11393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4324531846932721e-05,
+      "loss": 0.1509,
+      "step": 11394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4398941668097943e-05,
+      "loss": 3.47,
+      "step": 11395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4473529262269802e-05,
+      "loss": 1.3466,
+      "step": 11396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4548294474547242e-05,
+      "loss": 2.2729,
+      "step": 11397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4623237149661129e-05,
+      "loss": 2.6873,
+      "step": 11398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4698357131972984e-05,
+      "loss": 2.1307,
+      "step": 11399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4773654265476888e-05,
+      "loss": 0.4188,
+      "step": 11400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.484912839379824e-05,
+      "loss": 2.6884,
+      "step": 11401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.492477936019565e-05,
+      "loss": 2.539,
+      "step": 11402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5000607007559875e-05,
+      "loss": 2.4318,
+      "step": 11403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5076611178414779e-05,
+      "loss": 1.2663,
+      "step": 11404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.515279171491818e-05,
+      "loss": 1.143,
+      "step": 11405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5229148458860876e-05,
+      "loss": 1.4008,
+      "step": 11406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.530568125166847e-05,
+      "loss": 3.0848,
+      "step": 11407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5382389934400208e-05,
+      "loss": 1.5358,
+      "step": 11408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5459274347750808e-05,
+      "loss": 1.2108,
+      "step": 11409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5536334332049284e-05,
+      "loss": 2.2319,
+      "step": 11410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.561356972726058e-05,
+      "loss": 0.8402,
+      "step": 11411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5690980372985386e-05,
+      "loss": 2.4005,
+      "step": 11412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.576856610845985e-05,
+      "loss": 3.6422,
+      "step": 11413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5846326772557252e-05,
+      "loss": 0.3837,
+      "step": 11414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5924262203786842e-05,
+      "loss": 0.8246,
+      "step": 11415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6002372240295493e-05,
+      "loss": 1.4034,
+      "step": 11416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6080656719867437e-05,
+      "loss": 1.8488,
+      "step": 11417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6159115479924104e-05,
+      "loss": 0.2549,
+      "step": 11418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6237748357525695e-05,
+      "loss": 2.3444,
+      "step": 11419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6316555189370088e-05,
+      "loss": 0.6993,
+      "step": 11420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.639553581179465e-05,
+      "loss": 1.1839,
+      "step": 11421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6474690060775044e-05,
+      "loss": 1.1753,
+      "step": 11422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6554017771927153e-05,
+      "loss": 2.453,
+      "step": 11423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.663351878050582e-05,
+      "loss": 5.0266,
+      "step": 11424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6713192921406796e-05,
+      "loss": 3.1001,
+      "step": 11425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.679304002916545e-05,
+      "loss": 2.3366,
+      "step": 11426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.687305993795878e-05,
+      "loss": 1.1733,
+      "step": 11427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6953252481604087e-05,
+      "loss": 0.9099,
+      "step": 11428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.703361749356096e-05,
+      "loss": 1.8383,
+      "step": 11429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7114154806930027e-05,
+      "loss": 1.5408,
+      "step": 11430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7194864254454855e-05,
+      "loss": 2.3229,
+      "step": 11431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7275745668520758e-05,
+      "loss": 3.0047,
+      "step": 11432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.735679888115671e-05,
+      "loss": 1.2481,
+      "step": 11433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.743802372403409e-05,
+      "loss": 1.6275,
+      "step": 11434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7519420028468657e-05,
+      "loss": 2.4122,
+      "step": 11435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7600987625419463e-05,
+      "loss": 0.2114,
+      "step": 11436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7682726345489842e-05,
+      "loss": 1.4708,
+      "step": 11437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.776463601892836e-05,
+      "loss": 2.6318,
+      "step": 11438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7846716475627723e-05,
+      "loss": 2.9953,
+      "step": 11439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7928967545126783e-05,
+      "loss": 2.3226,
+      "step": 11440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.801138905660925e-05,
+      "loss": 1.7463,
+      "step": 11441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.809398083890549e-05,
+      "loss": 1.8854,
+      "step": 11442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8176742720492246e-05,
+      "loss": 2.8046,
+      "step": 11443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.825967452949241e-05,
+      "loss": 2.4296,
+      "step": 11444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.834277609367675e-05,
+      "loss": 2.4406,
+      "step": 11445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.842604724046274e-05,
+      "loss": 2.6086,
+      "step": 11446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8509487796916214e-05,
+      "loss": 1.0201,
+      "step": 11447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.859309758975125e-05,
+      "loss": 0.9024,
+      "step": 11448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8676876445329793e-05,
+      "loss": 1.4427,
+      "step": 11449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.876082418966355e-05,
+      "loss": 0.1408,
+      "step": 11450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.884494064841261e-05,
+      "loss": 2.8555,
+      "step": 11451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8929225646887573e-05,
+      "loss": 3.1553,
+      "step": 11452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9013679010048145e-05,
+      "loss": 2.481,
+      "step": 11453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9098300562505222e-05,
+      "loss": 1.1583,
+      "step": 11454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9183090128519586e-05,
+      "loss": 3.2421,
+      "step": 11455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.926804753200392e-05,
+      "loss": 1.6751,
+      "step": 11456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.935317259652153e-05,
+      "loss": 1.454,
+      "step": 11457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.943846514528833e-05,
+      "loss": 0.2199,
+      "step": 11458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.952392500117163e-05,
+      "loss": 3.2152,
+      "step": 11459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9609551986692053e-05,
+      "loss": 4.1835,
+      "step": 11460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9695345924022348e-05,
+      "loss": 1.7495,
+      "step": 11461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9781306634989348e-05,
+      "loss": 1.9754,
+      "step": 11462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9867433941072733e-05,
+      "loss": 2.901,
+      "step": 11463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.9953727663407018e-05,
+      "loss": 2.3467,
+      "step": 11464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.0040187622780273e-05,
+      "loss": 1.3383,
+      "step": 11465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.012681363963619e-05,
+      "loss": 0.4802,
+      "step": 11466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.0213605534072955e-05,
+      "loss": 2.6704,
+      "step": 11467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.0300563125844253e-05,
+      "loss": 1.135,
+      "step": 11468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.0387686234360315e-05,
+      "loss": 1.8813,
+      "step": 11469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.0474974678686777e-05,
+      "loss": 2.4597,
+      "step": 11470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.0562428277546752e-05,
+      "loss": 2.9299,
+      "step": 11471
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.0650046849319526e-05,
+      "loss": 1.8518,
+      "step": 11472
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.0737830212042632e-05,
+      "loss": 3.497,
+      "step": 11473
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.0825778183410493e-05,
+      "loss": 2.0256,
+      "step": 11474
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.0913890580776352e-05,
+      "loss": 2.9596,
+      "step": 11475
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1002167221151958e-05,
+      "loss": 1.2699,
+      "step": 11476
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1090607921207318e-05,
+      "loss": 0.6727,
+      "step": 11477
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1179212497272404e-05,
+      "loss": 2.9383,
+      "step": 11478
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1267980765336825e-05,
+      "loss": 0.1324,
+      "step": 11479
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1356912541049588e-05,
+      "loss": 1.9344,
+      "step": 11480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1446007639721034e-05,
+      "loss": 1.4949,
+      "step": 11481
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.153526587632142e-05,
+      "loss": 1.7605,
+      "step": 11482
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1624687065483095e-05,
+      "loss": 1.5855,
+      "step": 11483
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.171427102149911e-05,
+      "loss": 0.8517,
+      "step": 11484
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1804017558325362e-05,
+      "loss": 1.1081,
+      "step": 11485
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.189392648957924e-05,
+      "loss": 3.0583,
+      "step": 11486
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.1983997628541754e-05,
+      "loss": 3.5953,
+      "step": 11487
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.207423078815616e-05,
+      "loss": 1.8204,
+      "step": 11488
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.2164625781030124e-05,
+      "loss": 4.0293,
+      "step": 11489
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.2255182419434338e-05,
+      "loss": 3.561,
+      "step": 11490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.2345900515304673e-05,
+      "loss": 2.0184,
+      "step": 11491
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.243677988024082e-05,
+      "loss": 1.9368,
+      "step": 11492
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.252782032550842e-05,
+      "loss": 3.1225,
+      "step": 11493
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.261902166203772e-05,
+      "loss": 3.7455,
+      "step": 11494
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.271038370042572e-05,
+      "loss": 0.2008,
+      "step": 11495
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.2801906250934767e-05,
+      "loss": 1.9634,
+      "step": 11496
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.2893589123494785e-05,
+      "loss": 0.9572,
+      "step": 11497
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.2985432127701844e-05,
+      "loss": 0.1251,
+      "step": 11498
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.307743507282034e-05,
+      "loss": 0.8861,
+      "step": 11499
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.316959776778186e-05,
+      "loss": 4.0096,
+      "step": 11500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.326192002118621e-05,
+      "loss": 1.0442,
+      "step": 11501
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.3354401641302516e-05,
+      "loss": 2.6649,
+      "step": 11502
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.3447042436068024e-05,
+      "loss": 1.3476,
+      "step": 11503
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.3539842213090334e-05,
+      "loss": 2.9957,
+      "step": 11504
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.363280077964595e-05,
+      "loss": 1.0663,
+      "step": 11505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.372591794268232e-05,
+      "loss": 3.4827,
+      "step": 11506
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.3819193508817484e-05,
+      "loss": 3.3131,
+      "step": 11507
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.3912627284339873e-05,
+      "loss": 1.312,
+      "step": 11508
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.4006219075210013e-05,
+      "loss": 3.1233,
+      "step": 11509
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.409996868706028e-05,
+      "loss": 0.328,
+      "step": 11510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.419387592519461e-05,
+      "loss": 2.9316,
+      "step": 11511
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.4287940594590486e-05,
+      "loss": 3.5772,
+      "step": 11512
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.4382162499897556e-05,
+      "loss": 1.448,
+      "step": 11513
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.4476541445439805e-05,
+      "loss": 0.7535,
+      "step": 11514
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.4571077235214202e-05,
+      "loss": 1.9753,
+      "step": 11515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.4665769672892892e-05,
+      "loss": 2.0391,
+      "step": 11516
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.476061856182178e-05,
+      "loss": 2.2822,
+      "step": 11517
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.4855623705022767e-05,
+      "loss": 3.0217,
+      "step": 11518
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.4950784905192348e-05,
+      "loss": 0.3988,
+      "step": 11519
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.504610196470385e-05,
+      "loss": 1.3175,
+      "step": 11520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5141574685605963e-05,
+      "loss": 1.2538,
+      "step": 11521
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5237202869625076e-05,
+      "loss": 0.469,
+      "step": 11522
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5332986318163754e-05,
+      "loss": 1.5507,
+      "step": 11523
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5428924832303057e-05,
+      "loss": 1.7691,
+      "step": 11524
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5525018212801076e-05,
+      "loss": 1.2957,
+      "step": 11525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5621266260095234e-05,
+      "loss": 0.2355,
+      "step": 11526
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5717668774300785e-05,
+      "loss": 2.2269,
+      "step": 11527
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5814225555213145e-05,
+      "loss": 3.1558,
+      "step": 11528
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.5910936402306397e-05,
+      "loss": 2.1777,
+      "step": 11529
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.6007801114735642e-05,
+      "loss": 1.705,
+      "step": 11530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.6104819491335698e-05,
+      "loss": 2.404,
+      "step": 11531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.620199133062228e-05,
+      "loss": 1.1097,
+      "step": 11532
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.6299316430793088e-05,
+      "loss": 3.078,
+      "step": 11533
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.6396794589726603e-05,
+      "loss": 1.8855,
+      "step": 11534
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.6494425604984396e-05,
+      "loss": 1.2858,
+      "step": 11535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.659220927380962e-05,
+      "loss": 2.4782,
+      "step": 11536
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.669014539312913e-05,
+      "loss": 2.8411,
+      "step": 11537
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.6788233759553172e-05,
+      "loss": 1.116,
+      "step": 11538
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.688647416937511e-05,
+      "loss": 2.0802,
+      "step": 11539
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.6984866418573474e-05,
+      "loss": 2.1303,
+      "step": 11540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.708341030281053e-05,
+      "loss": 3.6558,
+      "step": 11541
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.718210561743434e-05,
+      "loss": 0.622,
+      "step": 11542
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.7280952157478513e-05,
+      "loss": 0.3211,
+      "step": 11543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.7379949717661813e-05,
+      "loss": 0.3547,
+      "step": 11544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.7479098092390397e-05,
+      "loss": 2.6903,
+      "step": 11545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.7578397075756223e-05,
+      "loss": 1.0526,
+      "step": 11546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.7677846461539448e-05,
+      "loss": 3.5602,
+      "step": 11547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.7777446043206878e-05,
+      "loss": 0.1668,
+      "step": 11548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.787719561391442e-05,
+      "loss": 1.5683,
+      "step": 11549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.7977094966505478e-05,
+      "loss": 0.5655,
+      "step": 11550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8077143893513393e-05,
+      "loss": 1.0134,
+      "step": 11551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8177342187159872e-05,
+      "loss": 1.6172,
+      "step": 11552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.827768963935744e-05,
+      "loss": 2.8941,
+      "step": 11553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.837818604170783e-05,
+      "loss": 4.2343,
+      "step": 11554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.847883118550445e-05,
+      "loss": 1.3187,
+      "step": 11555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8579624861730835e-05,
+      "loss": 2.0508,
+      "step": 11556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.868056686106302e-05,
+      "loss": 0.6836,
+      "step": 11557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.878165697386801e-05,
+      "loss": 0.2583,
+      "step": 11558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8882894990206223e-05,
+      "loss": 2.702,
+      "step": 11559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8984280699829925e-05,
+      "loss": 0.1337,
+      "step": 11560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9085813892185642e-05,
+      "loss": 1.4283,
+      "step": 11561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9187494356412624e-05,
+      "loss": 3.1409,
+      "step": 11562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9289321881345268e-05,
+      "loss": 0.7974,
+      "step": 11563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9391296255511812e-05,
+      "loss": 1.3286,
+      "step": 11564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9493417267135527e-05,
+      "loss": 1.3096,
+      "step": 11565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9595684704135885e-05,
+      "loss": 1.094,
+      "step": 11566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.969809835412729e-05,
+      "loss": 1.8265,
+      "step": 11567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.98006580044215e-05,
+      "loss": 5.6256,
+      "step": 11568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9903363442026032e-05,
+      "loss": 1.8356,
+      "step": 11569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.000621445364643e-05,
+      "loss": 1.9869,
+      "step": 11570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0109210825685874e-05,
+      "loss": 0.2897,
+      "step": 11571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.021235234424494e-05,
+      "loss": 4.8375,
+      "step": 11572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0315638795123514e-05,
+      "loss": 0.9969,
+      "step": 11573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.04190699638205e-05,
+      "loss": 1.2312,
+      "step": 11574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.05226456355335e-05,
+      "loss": 1.9266,
+      "step": 11575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.062636559516101e-05,
+      "loss": 2.0074,
+      "step": 11576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0730229627300876e-05,
+      "loss": 2.2625,
+      "step": 11577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.083423751625276e-05,
+      "loss": 1.066,
+      "step": 11578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.093838904601655e-05,
+      "loss": 1.7612,
+      "step": 11579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.104268400029482e-05,
+      "loss": 1.2886,
+      "step": 11580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.11471221624913e-05,
+      "loss": 1.547,
+      "step": 11581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.125170331571331e-05,
+      "loss": 1.4844,
+      "step": 11582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1356427242770204e-05,
+      "loss": 2.98,
+      "step": 11583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1461293726175844e-05,
+      "loss": 0.6963,
+      "step": 11584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1566302548147e-05,
+      "loss": 1.4717,
+      "step": 11585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.167145349060588e-05,
+      "loss": 0.2628,
+      "step": 11586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.177674633517853e-05,
+      "loss": 2.8983,
+      "step": 11587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.188218086319729e-05,
+      "loss": 0.1904,
+      "step": 11588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.1987756855699245e-05,
+      "loss": 2.5497,
+      "step": 11589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2093474093428746e-05,
+      "loss": 1.6693,
+      "step": 11590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.219933235683573e-05,
+      "loss": 1.7534,
+      "step": 11591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2305331426078344e-05,
+      "loss": 1.5077,
+      "step": 11592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2411471081021236e-05,
+      "loss": 0.5465,
+      "step": 11593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.2517751101238145e-05,
+      "loss": 1.4233,
+      "step": 11594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.262417126601028e-05,
+      "loss": 4.4425,
+      "step": 11595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.273073135432884e-05,
+      "loss": 1.4515,
+      "step": 11596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.28374311448936e-05,
+      "loss": 0.8233,
+      "step": 11597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.29442704161143e-05,
+      "loss": 0.8171,
+      "step": 11598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3051248946111724e-05,
+      "loss": 1.5452,
+      "step": 11599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3158366512716434e-05,
+      "loss": 0.2791,
+      "step": 11600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3265622893471016e-05,
+      "loss": 0.2824,
+      "step": 11601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3373017865629786e-05,
+      "loss": 4.3232,
+      "step": 11602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.348055120615845e-05,
+      "loss": 2.3382,
+      "step": 11603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3588222691736116e-05,
+      "loss": 3.1594,
+      "step": 11604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.369603209875505e-05,
+      "loss": 1.2143,
+      "step": 11605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.380397920332026e-05,
+      "loss": 1.3956,
+      "step": 11606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.3912063781251836e-05,
+      "loss": 2.3046,
+      "step": 11607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.402028560808335e-05,
+      "loss": 0.6602,
+      "step": 11608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4128644459064364e-05,
+      "loss": 3.0579,
+      "step": 11609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.423714010915883e-05,
+      "loss": 2.2742,
+      "step": 11610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4345772333047665e-05,
+      "loss": 2.9451,
+      "step": 11611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4454540905127084e-05,
+      "loss": 0.7767,
+      "step": 11612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.45634455995112e-05,
+      "loss": 0.1153,
+      "step": 11613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.4672486190030384e-05,
+      "loss": 3.4441,
+      "step": 11614
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.478166245023383e-05,
+      "loss": 2.0923,
+      "step": 11615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.489097415338791e-05,
+      "loss": 2.7027,
+      "step": 11616
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.5000421072478816e-05,
+      "loss": 0.9665,
+      "step": 11617
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.511000298021082e-05,
+      "loss": 1.3259,
+      "step": 11618
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.521971964900893e-05,
+      "loss": 1.4708,
+      "step": 11619
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.532957085101725e-05,
+      "loss": 0.2452,
+      "step": 11620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.543955635810151e-05,
+      "loss": 3.6596,
+      "step": 11621
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.554967594184749e-05,
+      "loss": 0.7589,
+      "step": 11622
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.565992937356359e-05,
+      "loss": 0.545,
+      "step": 11623
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.577031642427916e-05,
+      "loss": 1.5689,
+      "step": 11624
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.588083686474711e-05,
+      "loss": 0.4798,
+      "step": 11625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.599149046544229e-05,
+      "loss": 0.092,
+      "step": 11626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.610227699656407e-05,
+      "loss": 2.0136,
+      "step": 11627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.621319622803494e-05,
+      "loss": 0.2184,
+      "step": 11628
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.632424792950181e-05,
+      "loss": 0.402,
+      "step": 11629
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.64354318703373e-05,
+      "loss": 1.3654,
+      "step": 11630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.6546747819638305e-05,
+      "loss": 0.7909,
+      "step": 11631
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.665819554622838e-05,
+      "loss": 0.6181,
+      "step": 11632
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.676977481865742e-05,
+      "loss": 1.7164,
+      "step": 11633
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.688148540520132e-05,
+      "loss": 0.8093,
+      "step": 11634
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.699332707386407e-05,
+      "loss": 1.9462,
+      "step": 11635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.710529959237744e-05,
+      "loss": 2.5019,
+      "step": 11636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7217402728200634e-05,
+      "loss": 1.0086,
+      "step": 11637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.732963624852269e-05,
+      "loss": 0.3967,
+      "step": 11638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7441999920260784e-05,
+      "loss": 0.2734,
+      "step": 11639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.755449351006288e-05,
+      "loss": 2.6005,
+      "step": 11640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7667116784306045e-05,
+      "loss": 1.0165,
+      "step": 11641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.7779869509099106e-05,
+      "loss": 0.8855,
+      "step": 11642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.789275145028097e-05,
+      "loss": 3.1177,
+      "step": 11643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.800576237342328e-05,
+      "loss": 1.0026,
+      "step": 11644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.811890204382873e-05,
+      "loss": 0.6574,
+      "step": 11645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8232170226533725e-05,
+      "loss": 0.0908,
+      "step": 11646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.834556668630668e-05,
+      "loss": 1.3402,
+      "step": 11647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.845909118765071e-05,
+      "loss": 1.4421,
+      "step": 11648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.857274349480191e-05,
+      "loss": 2.7206,
+      "step": 11649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8686523371732084e-05,
+      "loss": 1.8865,
+      "step": 11650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8800430582146965e-05,
+      "loss": 0.8321,
+      "step": 11651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8914464889488966e-05,
+      "loss": 0.0825,
+      "step": 11652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.902862605693543e-05,
+      "loss": 1.2742,
+      "step": 11653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.9142913847401396e-05,
+      "loss": 2.052,
+      "step": 11654
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.925732802353776e-05,
+      "loss": 1.0177,
+      "step": 11655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.937186834773411e-05,
+      "loss": 4.1816,
+      "step": 11656
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.948653458211691e-05,
+      "loss": 1.777,
+      "step": 11657
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.960132648855227e-05,
+      "loss": 2.9687,
+      "step": 11658
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.971624382864418e-05,
+      "loss": 2.2126,
+      "step": 11659
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.9831286363737273e-05,
+      "loss": 2.0351,
+      "step": 11660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.994645385491532e-05,
+      "loss": 1.6412,
+      "step": 11661
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.0061746063002604e-05,
+      "loss": 0.9267,
+      "step": 11662
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.0177162748564935e-05,
+      "loss": 3.8697,
+      "step": 11663
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.029270367190965e-05,
+      "loss": 1.9926,
+      "step": 11664
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.04083685930852e-05,
+      "loss": 3.5913,
+      "step": 11665
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.052415727188368e-05,
+      "loss": 1.4922,
+      "step": 11666
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.0640069467839e-05,
+      "loss": 0.135,
+      "step": 11667
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.075610494022945e-05,
+      "loss": 2.1536,
+      "step": 11668
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.087226344807724e-05,
+      "loss": 4.0717,
+      "step": 11669
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.098854475014829e-05,
+      "loss": 1.7385,
+      "step": 11670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.110494860495459e-05,
+      "loss": 1.3812,
+      "step": 11671
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.122147477075249e-05,
+      "loss": 3.5792,
+      "step": 11672
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.1338123005545483e-05,
+      "loss": 0.1901,
+      "step": 11673
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.145489306708239e-05,
+      "loss": 4.1228,
+      "step": 11674
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.1571784712860194e-05,
+      "loss": 0.9412,
+      "step": 11675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.168879770012221e-05,
+      "loss": 0.4027,
+      "step": 11676
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.1805931785860916e-05,
+      "loss": 0.5292,
+      "step": 11677
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.192318672681614e-05,
+      "loss": 0.6367,
+      "step": 11678
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.204056227947788e-05,
+      "loss": 0.698,
+      "step": 11679
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.215805820008451e-05,
+      "loss": 1.7089,
+      "step": 11680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.227567424462555e-05,
+      "loss": 1.039,
+      "step": 11681
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.239341016883992e-05,
+      "loss": 0.1895,
+      "step": 11682
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.251126572821874e-05,
+      "loss": 1.3006,
+      "step": 11683
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.26292406780035e-05,
+      "loss": 2.0238,
+      "step": 11684
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.274733477318893e-05,
+      "loss": 2.8467,
+      "step": 11685
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.286554776852112e-05,
+      "loss": 2.6729,
+      "step": 11686
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.298387941850044e-05,
+      "loss": 1.1576,
+      "step": 11687
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.310232947737965e-05,
+      "loss": 2.5532,
+      "step": 11688
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.322089769916678e-05,
+      "loss": 2.4139,
+      "step": 11689
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.33395838376233e-05,
+      "loss": 1.4191,
+      "step": 11690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.345838764626695e-05,
+      "loss": 1.5921,
+      "step": 11691
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.3577308878370234e-05,
+      "loss": 1.0305,
+      "step": 11692
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.369634728696179e-05,
+      "loss": 0.384,
+      "step": 11693
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.3815502624827785e-05,
+      "loss": 3.6845,
+      "step": 11694
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.3934774644510365e-05,
+      "loss": 1.9196,
+      "step": 11695
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.4054163098310284e-05,
+      "loss": 0.6569,
+      "step": 11696
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.417366773828643e-05,
+      "loss": 1.6597,
+      "step": 11697
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.429328831625555e-05,
+      "loss": 0.3424,
+      "step": 11698
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.44130245837945e-05,
+      "loss": 2.8941,
+      "step": 11699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.4532876292239875e-05,
+      "loss": 2.8056,
+      "step": 11700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.4652843192687634e-05,
+      "loss": 1.6645,
+      "step": 11701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.477292503599568e-05,
+      "loss": 0.6004,
+      "step": 11702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.489312157278206e-05,
+      "loss": 3.6832,
+      "step": 11703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.501343255342777e-05,
+      "loss": 2.45,
+      "step": 11704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.513385772807499e-05,
+      "loss": 2.5621,
+      "step": 11705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.525439684662991e-05,
+      "loss": 1.5385,
+      "step": 11706
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.5375049658760894e-05,
+      "loss": 2.0497,
+      "step": 11707
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.5495815913901394e-05,
+      "loss": 5.0595,
+      "step": 11708
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.5616695361248066e-05,
+      "loss": 3.2102,
+      "step": 11709
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.573768774976368e-05,
+      "loss": 2.0343,
+      "step": 11710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.5858792828175304e-05,
+      "loss": 4.9565,
+      "step": 11711
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.5980010344977096e-05,
+      "loss": 0.0907,
+      "step": 11712
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.6101340048428575e-05,
+      "loss": 0.1831,
+      "step": 11713
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.622278168655745e-05,
+      "loss": 1.9586,
+      "step": 11714
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.6344335007157766e-05,
+      "loss": 2.1098,
+      "step": 11715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.6465999757792825e-05,
+      "loss": 2.3715,
+      "step": 11716
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.658777568579332e-05,
+      "loss": 0.5349,
+      "step": 11717
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.670966253826028e-05,
+      "loss": 3.1128,
+      "step": 11718
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.683166006206312e-05,
+      "loss": 1.7069,
+      "step": 11719
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.695376800384267e-05,
+      "loss": 1.3557,
+      "step": 11720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.7075986110009207e-05,
+      "loss": 0.4574,
+      "step": 11721
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.7198314126745446e-05,
+      "loss": 2.8224,
+      "step": 11722
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.732075180000462e-05,
+      "loss": 2.7073,
+      "step": 11723
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.744329887551349e-05,
+      "loss": 2.1962,
+      "step": 11724
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.756595509877062e-05,
+      "loss": 0.8096,
+      "step": 11725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.768872021504799e-05,
+      "loss": 3.1612,
+      "step": 11726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.781159396939199e-05,
+      "loss": 3.8314,
+      "step": 11727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.79345761066234e-05,
+      "loss": 1.0513,
+      "step": 11728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.805766637133701e-05,
+      "loss": 2.3321,
+      "step": 11729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.818086450790398e-05,
+      "loss": 0.666,
+      "step": 11730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.83041702604714e-05,
+      "loss": 1.6882,
+      "step": 11731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.8427583372962034e-05,
+      "loss": 2.5119,
+      "step": 11732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.855110358907683e-05,
+      "loss": 1.7281,
+      "step": 11733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.867473065229311e-05,
+      "loss": 3.8537,
+      "step": 11734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.879846430586753e-05,
+      "loss": 1.2462,
+      "step": 11735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.892230429283415e-05,
+      "loss": 2.4843,
+      "step": 11736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.904625035600745e-05,
+      "loss": 2.1796,
+      "step": 11737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.917030223798035e-05,
+      "loss": 1.1043,
+      "step": 11738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.929445968112726e-05,
+      "loss": 2.4686,
+      "step": 11739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.9418722427602136e-05,
+      "loss": 1.178,
+      "step": 11740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.954309021934147e-05,
+      "loss": 1.7495,
+      "step": 11741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.9667562798062376e-05,
+      "loss": 0.32,
+      "step": 11742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.9792139905265555e-05,
+      "loss": 0.453,
+      "step": 11743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.991682128223344e-05,
+      "loss": 2.0417,
+      "step": 11744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.0041606670033116e-05,
+      "loss": 2.0704,
+      "step": 11745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.0166495809514444e-05,
+      "loss": 3.7226,
+      "step": 11746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.029148844131304e-05,
+      "loss": 2.5939,
+      "step": 11747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.041658430584839e-05,
+      "loss": 3.4227,
+      "step": 11748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.054178314332677e-05,
+      "loss": 0.8883,
+      "step": 11749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.066708469373942e-05,
+      "loss": 2.6033,
+      "step": 11750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.0792488696865504e-05,
+      "loss": 2.9829,
+      "step": 11751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.091799489227012e-05,
+      "loss": 2.9194,
+      "step": 11752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.1043603019307454e-05,
+      "loss": 3.1391,
+      "step": 11753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.1169312817118696e-05,
+      "loss": 1.1044,
+      "step": 11754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.129512402463521e-05,
+      "loss": 2.0477,
+      "step": 11755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.142103638057673e-05,
+      "loss": 2.7108,
+      "step": 11756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.1547049623453004e-05,
+      "loss": 1.1617,
+      "step": 11757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.1673163491564835e-05,
+      "loss": 1.1591,
+      "step": 11758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.1799377723004096e-05,
+      "loss": 1.4132,
+      "step": 11759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.192569205565325e-05,
+      "loss": 2.5948,
+      "step": 11760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.205210622718817e-05,
+      "loss": 0.6331,
+      "step": 11761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.217861997507608e-05,
+      "loss": 2.3721,
+      "step": 11762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.230523303657839e-05,
+      "loss": 3.9172,
+      "step": 11763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.243194514875027e-05,
+      "loss": 1.9878,
+      "step": 11764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.255875604844018e-05,
+      "loss": 1.2901,
+      "step": 11765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.268566547229273e-05,
+      "loss": 0.7016,
+      "step": 11766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.2812673156746675e-05,
+      "loss": 0.9931,
+      "step": 11767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.29397788380379e-05,
+      "loss": 0.1206,
+      "step": 11768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.3066982252197585e-05,
+      "loss": 0.5827,
+      "step": 11769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.3194283135055155e-05,
+      "loss": 2.6776,
+      "step": 11770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.3321681222236376e-05,
+      "loss": 1.9008,
+      "step": 11771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.344917624916641e-05,
+      "loss": 0.1244,
+      "step": 11772
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.3576767951067805e-05,
+      "loss": 2.3217,
+      "step": 11773
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.370445606296364e-05,
+      "loss": 1.7594,
+      "step": 11774
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.3832240319675485e-05,
+      "loss": 0.9932,
+      "step": 11775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.3960120455826525e-05,
+      "loss": 0.2639,
+      "step": 11776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.408809620583952e-05,
+      "loss": 1.09,
+      "step": 11777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.421616730393998e-05,
+      "loss": 2.7113,
+      "step": 11778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.4344333484154096e-05,
+      "loss": 0.173,
+      "step": 11779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.447259448031189e-05,
+      "loss": 0.1466,
+      "step": 11780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.4600950026045183e-05,
+      "loss": 0.6761,
+      "step": 11781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.472939985479073e-05,
+      "loss": 3.478,
+      "step": 11782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.485794369978818e-05,
+      "loss": 2.1315,
+      "step": 11783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.498658129408323e-05,
+      "loss": 1.2577,
+      "step": 11784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.511531237052563e-05,
+      "loss": 2.8997,
+      "step": 11785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.524413666177219e-05,
+      "loss": 0.7229,
+      "step": 11786
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.537305390028495e-05,
+      "loss": 4.8873,
+      "step": 11787
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.550206381833415e-05,
+      "loss": 1.8093,
+      "step": 11788
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.5631166147996264e-05,
+      "loss": 0.5028,
+      "step": 11789
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.576036062115715e-05,
+      "loss": 0.8593,
+      "step": 11790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.5889646969510036e-05,
+      "loss": 3.0998,
+      "step": 11791
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.601902492455859e-05,
+      "loss": 2.0946,
+      "step": 11792
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.614849421761503e-05,
+      "loss": 0.2007,
+      "step": 11793
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.627805457980274e-05,
+      "loss": 0.6371,
+      "step": 11794
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6407705742056074e-05,
+      "loss": 2.343,
+      "step": 11795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6537447435119775e-05,
+      "loss": 0.4376,
+      "step": 11796
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.666727938955192e-05,
+      "loss": 0.6044,
+      "step": 11797
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.679720133572183e-05,
+      "loss": 2.2738,
+      "step": 11798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.6927213003813195e-05,
+      "loss": 2.562,
+      "step": 11799
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7057314123822114e-05,
+      "loss": 2.4343,
+      "step": 11800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7187504425560214e-05,
+      "loss": 0.1647,
+      "step": 11801
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.731778363865259e-05,
+      "loss": 3.1597,
+      "step": 11802
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7448151492541e-05,
+      "loss": 0.8749,
+      "step": 11803
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.757860771648182e-05,
+      "loss": 2.8962,
+      "step": 11804
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.770915203954921e-05,
+      "loss": 1.8876,
+      "step": 11805
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.7839784190633036e-05,
+      "loss": 0.4608,
+      "step": 11806
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.797050389844212e-05,
+      "loss": 4.615,
+      "step": 11807
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.810131089150208e-05,
+      "loss": 0.1284,
+      "step": 11808
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8232204898158635e-05,
+      "loss": 1.8519,
+      "step": 11809
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.836318564657546e-05,
+      "loss": 1.4507,
+      "step": 11810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8494252864737374e-05,
+      "loss": 0.1567,
+      "step": 11811
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.862540628044836e-05,
+      "loss": 1.7408,
+      "step": 11812
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.875664562133465e-05,
+      "loss": 0.9518,
+      "step": 11813
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.8887970614842745e-05,
+      "loss": 0.2069,
+      "step": 11814
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.901938098824259e-05,
+      "loss": 1.0757,
+      "step": 11815
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.915087646862547e-05,
+      "loss": 2.673,
+      "step": 11816
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.928245678290725e-05,
+      "loss": 3.581,
+      "step": 11817
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.941412165782633e-05,
+      "loss": 3.6106,
+      "step": 11818
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.954587081994676e-05,
+      "loss": 2.8258,
+      "step": 11819
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9677703995656595e-05,
+      "loss": 4.2636,
+      "step": 11820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.980962091116942e-05,
+      "loss": 3.2545,
+      "step": 11821
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.994162129252553e-05,
+      "loss": 1.0571,
+      "step": 11822
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.007370486559192e-05,
+      "loss": 1.7812,
+      "step": 11823
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.020587135606182e-05,
+      "loss": 0.2738,
+      "step": 11824
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.033812048945724e-05,
+      "loss": 1.4403,
+      "step": 11825
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0470451991128555e-05,
+      "loss": 1.5194,
+      "step": 11826
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060286558625411e-05,
+      "loss": 2.6322,
+      "step": 11827
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.073536099984307e-05,
+      "loss": 0.1344,
+      "step": 11828
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.086793795673336e-05,
+      "loss": 2.6576,
+      "step": 11829
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.100059618159488e-05,
+      "loss": 1.9588,
+      "step": 11830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.113333539892745e-05,
+      "loss": 3.7486,
+      "step": 11831
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.126615533306398e-05,
+      "loss": 4.3506,
+      "step": 11832
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.139905570816847e-05,
+      "loss": 2.068,
+      "step": 11833
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.153203624823913e-05,
+      "loss": 3.5722,
+      "step": 11834
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.166509667710642e-05,
+      "loss": 3.1303,
+      "step": 11835
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.179823671843619e-05,
+      "loss": 1.6762,
+      "step": 11836
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.19314560957276e-05,
+      "loss": 0.306,
+      "step": 11837
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.206475453231644e-05,
+      "loss": 0.7795,
+      "step": 11838
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.219813175137292e-05,
+      "loss": 1.2087,
+      "step": 11839
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.2331587475905e-05,
+      "loss": 1.1024,
+      "step": 11840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.246512142875626e-05,
+      "loss": 3.0559,
+      "step": 11841
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.259873333260917e-05,
+      "loss": 0.3741,
+      "step": 11842
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.273242290998296e-05,
+      "loss": 1.048,
+      "step": 11843
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.286618988323689e-05,
+      "loss": 3.9151,
+      "step": 11844
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.300003397456817e-05,
+      "loss": 2.9035,
+      "step": 11845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.313395490601517e-05,
+      "loss": 2.777,
+      "step": 11846
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.326795239945532e-05,
+      "loss": 1.735,
+      "step": 11847
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.340202617660845e-05,
+      "loss": 1.9098,
+      "step": 11848
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.353617595903458e-05,
+      "loss": 1.9473,
+      "step": 11849
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.367040146813724e-05,
+      "loss": 3.2367,
+      "step": 11850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.380470242516137e-05,
+      "loss": 2.3079,
+      "step": 11851
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.39390785511965e-05,
+      "loss": 1.1869,
+      "step": 11852
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.40735295671748e-05,
+      "loss": 2.9095,
+      "step": 11853
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.420805519387419e-05,
+      "loss": 4.2083,
+      "step": 11854
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.434265515191633e-05,
+      "loss": 2.0076,
+      "step": 11855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.44773291617695e-05,
+      "loss": 0.8242,
+      "step": 11856
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.461207694374819e-05,
+      "loss": 0.1854,
+      "step": 11857
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.47468982180127e-05,
+      "loss": 2.817,
+      "step": 11858
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.488179270457202e-05,
+      "loss": 1.4138,
+      "step": 11859
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.501676012328177e-05,
+      "loss": 1.9708,
+      "step": 11860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.515180019384746e-05,
+      "loss": 1.8339,
+      "step": 11861
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.528691263582231e-05,
+      "loss": 3.1679,
+      "step": 11862
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.542209716861071e-05,
+      "loss": 1.5622,
+      "step": 11863
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.555735351146585e-05,
+      "loss": 3.1154,
+      "step": 11864
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.569268138349324e-05,
+      "loss": 0.6834,
+      "step": 11865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.582808050364843e-05,
+      "loss": 1.7494,
+      "step": 11866
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.596355059074039e-05,
+      "loss": 0.3616,
+      "step": 11867
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.609909136342936e-05,
+      "loss": 1.1647,
+      "step": 11868
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.623470254023008e-05,
+      "loss": 1.1343,
+      "step": 11869
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.63703838395098e-05,
+      "loss": 1.9105,
+      "step": 11870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.650613497949144e-05,
+      "loss": 2.7583,
+      "step": 11871
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.664195567825156e-05,
+      "loss": 0.5939,
+      "step": 11872
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.677784565372358e-05,
+      "loss": 0.5706,
+      "step": 11873
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.691380462369572e-05,
+      "loss": 1.2172,
+      "step": 11874
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.704983230581426e-05,
+      "loss": 2.0396,
+      "step": 11875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.718592841758146e-05,
+      "loss": 0.311,
+      "step": 11876
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.732209267635876e-05,
+      "loss": 0.5537,
+      "step": 11877
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.745832479936478e-05,
+      "loss": 0.2097,
+      "step": 11878
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.759462450367852e-05,
+      "loss": 1.6047,
+      "step": 11879
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.773099150623728e-05,
+      "loss": 2.1822,
+      "step": 11880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.786742552383997e-05,
+      "loss": 3.2252,
+      "step": 11881
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.800392627314494e-05,
+      "loss": 3.702,
+      "step": 11882
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.814049347067335e-05,
+      "loss": 2.1703,
+      "step": 11883
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.827712683280698e-05,
+      "loss": 2.5913,
+      "step": 11884
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.841382607579155e-05,
+      "loss": 2.8472,
+      "step": 11885
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.855059091573464e-05,
+      "loss": 2.8616,
+      "step": 11886
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.868742106860891e-05,
+      "loss": 0.228,
+      "step": 11887
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.882431625025006e-05,
+      "loss": 3.0791,
+      "step": 11888
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.896127617635977e-05,
+      "loss": 0.7104,
+      "step": 11889
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.909830056250519e-05,
+      "loss": 2.5589,
+      "step": 11890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.92353891241187e-05,
+      "loss": 0.7925,
+      "step": 11891
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.937254157650063e-05,
+      "loss": 2.4869,
+      "step": 11892
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.950975763481735e-05,
+      "loss": 2.174,
+      "step": 11893
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.96470370141045e-05,
+      "loss": 3.3519,
+      "step": 11894
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.978437942926482e-05,
+      "loss": 0.8692,
+      "step": 11895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.992178459507147e-05,
+      "loss": 3.7343,
+      "step": 11896
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.005925222616603e-05,
+      "loss": 1.7968,
+      "step": 11897
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.019678203706162e-05,
+      "loss": 1.8702,
+      "step": 11898
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.033437374214091e-05,
+      "loss": 0.2716,
+      "step": 11899
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.04720270556594e-05,
+      "loss": 1.192,
+      "step": 11900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.060974169174327e-05,
+      "loss": 0.426,
+      "step": 11901
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.07475173643927e-05,
+      "loss": 0.2224,
+      "step": 11902
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.088535378747976e-05,
+      "loss": 0.827,
+      "step": 11903
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.102325067475171e-05,
+      "loss": 1.0354,
+      "step": 11904
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.116120773982887e-05,
+      "loss": 4.4284,
+      "step": 11905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.129922469620799e-05,
+      "loss": 3.8837,
+      "step": 11906
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.143730125726004e-05,
+      "loss": 3.4752,
+      "step": 11907
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.157543713623356e-05,
+      "loss": 0.5616,
+      "step": 11908
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.171363204625257e-05,
+      "loss": 0.4954,
+      "step": 11909
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.185188570031985e-05,
+      "loss": 0.4767,
+      "step": 11910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.19901978113148e-05,
+      "loss": 1.0951,
+      "step": 11911
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.21285680919968e-05,
+      "loss": 3.4458,
+      "step": 11912
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.226699625500304e-05,
+      "loss": 2.0622,
+      "step": 11913
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.240548201285191e-05,
+      "loss": 2.7551,
+      "step": 11914
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.25440250779408e-05,
+      "loss": 0.7661,
+      "step": 11915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.268262516254942e-05,
+      "loss": 4.3779,
+      "step": 11916
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.282128197883775e-05,
+      "loss": 0.6487,
+      "step": 11917
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.29599952388493e-05,
+      "loss": 1.3098,
+      "step": 11918
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.3098764654509e-05,
+      "loss": 2.6109,
+      "step": 11919
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.323758993762612e-05,
+      "loss": 1.7332,
+      "step": 11920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.337647079989402e-05,
+      "loss": 2.6377,
+      "step": 11921
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.351540695288952e-05,
+      "loss": 0.6358,
+      "step": 11922
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.365439810807602e-05,
+      "loss": 1.0525,
+      "step": 11923
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.37934439768013e-05,
+      "loss": 1.765,
+      "step": 11924
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.39325442703009e-05,
+      "loss": 2.0971,
+      "step": 11925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.407169869969599e-05,
+      "loss": 0.7476,
+      "step": 11926
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.421090697599664e-05,
+      "loss": 1.0939,
+      "step": 11927
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.435016881009973e-05,
+      "loss": 1.3438,
+      "step": 11928
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.44894839127923e-05,
+      "loss": 1.199,
+      "step": 11929
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.462885199474936e-05,
+      "loss": 2.5343,
+      "step": 11930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.47682727665373e-05,
+      "loss": 2.6069,
+      "step": 11931
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.490774593861173e-05,
+      "loss": 0.4302,
+      "step": 11932
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.504727122132077e-05,
+      "loss": 0.6624,
+      "step": 11933
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.518684832490294e-05,
+      "loss": 1.2349,
+      "step": 11934
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.532647695949054e-05,
+      "loss": 1.5172,
+      "step": 11935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.54661568351075e-05,
+      "loss": 0.6912,
+      "step": 11936
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.560588766167262e-05,
+      "loss": 2.5229,
+      "step": 11937
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.574566914899763e-05,
+      "loss": 1.6907,
+      "step": 11938
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.588550100679034e-05,
+      "loss": 1.3335,
+      "step": 11939
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.60253829446526e-05,
+      "loss": 0.3588,
+      "step": 11940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.616531467208365e-05,
+      "loss": 0.8963,
+      "step": 11941
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.630529589847794e-05,
+      "loss": 2.9283,
+      "step": 11942
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.644532633312849e-05,
+      "loss": 1.8284,
+      "step": 11943
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.658540568522474e-05,
+      "loss": 1.6201,
+      "step": 11944
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.672553366385597e-05,
+      "loss": 1.853,
+      "step": 11945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.686570997800902e-05,
+      "loss": 3.4667,
+      "step": 11946
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.700593433657179e-05,
+      "loss": 2.3504,
+      "step": 11947
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.7146206448331e-05,
+      "loss": 4.5685,
+      "step": 11948
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.728652602197559e-05,
+      "loss": 0.1052,
+      "step": 11949
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.742689276609452e-05,
+      "loss": 0.2873,
+      "step": 11950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.756730638917987e-05,
+      "loss": 1.9865,
+      "step": 11951
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.770776659962631e-05,
+      "loss": 4.3057,
+      "step": 11952
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.784827310573076e-05,
+      "loss": 2.4128,
+      "step": 11953
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.798882561569538e-05,
+      "loss": 1.0665,
+      "step": 11954
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.812942383762541e-05,
+      "loss": 0.495,
+      "step": 11955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.82700674795325e-05,
+      "loss": 0.4516,
+      "step": 11956
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.841075624933262e-05,
+      "loss": 0.2484,
+      "step": 11957
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.855148985484942e-05,
+      "loss": 1.8612,
+      "step": 11958
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.869226800381203e-05,
+      "loss": 0.1736,
+      "step": 11959
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.883309040385845e-05,
+      "loss": 1.0072,
+      "step": 11960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.897395676253343e-05,
+      "loss": 1.6341,
+      "step": 11961
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.911486678729177e-05,
+      "loss": 1.1618,
+      "step": 11962
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.925582018549623e-05,
+      "loss": 0.3215,
+      "step": 11963
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.939681666442089e-05,
+      "loss": 3.9532,
+      "step": 11964
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.953785593124893e-05,
+      "loss": 1.9824,
+      "step": 11965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.967893769307609e-05,
+      "loss": 2.404,
+      "step": 11966
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.982006165690848e-05,
+      "loss": 2.7574,
+      "step": 11967
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 7.996122752966594e-05,
+      "loss": 2.4468,
+      "step": 11968
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.010243501817987e-05,
+      "loss": 0.854,
+      "step": 11969
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.024368382919668e-05,
+      "loss": 0.3777,
+      "step": 11970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.038497366937548e-05,
+      "loss": 0.7272,
+      "step": 11971
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.05263042452917e-05,
+      "loss": 2.9186,
+      "step": 11972
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.066767526343465e-05,
+      "loss": 3.0851,
+      "step": 11973
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.080908643021112e-05,
+      "loss": 0.7357,
+      "step": 11974
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.095053745194311e-05,
+      "loss": 0.2669,
+      "step": 11975
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.109202803487125e-05,
+      "loss": 2.7535,
+      "step": 11976
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.123355788515258e-05,
+      "loss": 1.5765,
+      "step": 11977
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.137512670886402e-05,
+      "loss": 3.9441,
+      "step": 11978
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.151673421200018e-05,
+      "loss": 2.0308,
+      "step": 11979
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.165838010047667e-05,
+      "loss": 1.2197,
+      "step": 11980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.180006408012804e-05,
+      "loss": 0.9992,
+      "step": 11981
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.194178585671077e-05,
+      "loss": 2.865,
+      "step": 11982
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.208354513590284e-05,
+      "loss": 2.8822,
+      "step": 11983
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.222534162330327e-05,
+      "loss": 1.4911,
+      "step": 11984
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.236717502443527e-05,
+      "loss": 2.436,
+      "step": 11985
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.250904504474396e-05,
+      "loss": 3.1902,
+      "step": 11986
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.265095138959975e-05,
+      "loss": 0.9438,
+      "step": 11987
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.279289376429631e-05,
+      "loss": 0.2876,
+      "step": 11988
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.293487187405382e-05,
+      "loss": 2.7951,
+      "step": 11989
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.307688542401689e-05,
+      "loss": 2.1898,
+      "step": 11990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.32189341192579e-05,
+      "loss": 1.6761,
+      "step": 11991
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.336101766477484e-05,
+      "loss": 1.098,
+      "step": 11992
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.350313576549472e-05,
+      "loss": 0.2518,
+      "step": 11993
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.364528812627142e-05,
+      "loss": 0.1206,
+      "step": 11994
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.378747445188906e-05,
+      "loss": 0.7715,
+      "step": 11995
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.39296944470598e-05,
+      "loss": 0.8434,
+      "step": 11996
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.407194781642727e-05,
+      "loss": 1.8215,
+      "step": 11997
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.421423426456444e-05,
+      "loss": 1.868,
+      "step": 11998
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.435655349597695e-05,
+      "loss": 2.0066,
+      "step": 11999
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.449890521510092e-05,
+      "loss": 3.5868,
+      "step": 12000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.464128912630649e-05,
+      "loss": 1.8897,
+      "step": 12001
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.478370493389547e-05,
+      "loss": 1.8644,
+      "step": 12002
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.492615234210485e-05,
+      "loss": 2.6999,
+      "step": 12003
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.506863105510462e-05,
+      "loss": 2.4593,
+      "step": 12004
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.521114077700111e-05,
+      "loss": 0.8368,
+      "step": 12005
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.535368121183489e-05,
+      "loss": 4.7048,
+      "step": 12006
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.549625206358411e-05,
+      "loss": 2.0645,
+      "step": 12007
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.563885303616241e-05,
+      "loss": 0.6701,
+      "step": 12008
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.578148383342222e-05,
+      "loss": 1.5212,
+      "step": 12009
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.592414415915265e-05,
+      "loss": 0.7175,
+      "step": 12010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.606683371708288e-05,
+      "loss": 0.6159,
+      "step": 12011
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.620955221088001e-05,
+      "loss": 0.2645,
+      "step": 12012
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.63522993441524e-05,
+      "loss": 2.5659,
+      "step": 12013
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.649507482044759e-05,
+      "loss": 0.8984,
+      "step": 12014
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.663787834325526e-05,
+      "loss": 4.3491,
+      "step": 12015
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.678070961600686e-05,
+      "loss": 4.3497,
+      "step": 12016
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.692356834207515e-05,
+      "loss": 3.0827,
+      "step": 12017
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.706645422477733e-05,
+      "loss": 0.9521,
+      "step": 12018
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.720936696737273e-05,
+      "loss": 1.673,
+      "step": 12019
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.735230627306632e-05,
+      "loss": 2.0468,
+      "step": 12020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.749527184500652e-05,
+      "loss": 1.54,
+      "step": 12021
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.763826338628861e-05,
+      "loss": 1.4702,
+      "step": 12022
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.77812805999525e-05,
+      "loss": 0.1782,
+      "step": 12023
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.792432318898623e-05,
+      "loss": 3.4948,
+      "step": 12024
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.806739085632369e-05,
+      "loss": 3.1687,
+      "step": 12025
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.821048330484813e-05,
+      "loss": 2.4831,
+      "step": 12026
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.835360023738992e-05,
+      "loss": 2.4867,
+      "step": 12027
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.849674135672998e-05,
+      "loss": 1.4348,
+      "step": 12028
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.863990636559762e-05,
+      "loss": 0.996,
+      "step": 12029
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.878309496667391e-05,
+      "loss": 2.818,
+      "step": 12030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.892630686258953e-05,
+      "loss": 2.3611,
+      "step": 12031
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.90695417559282e-05,
+      "loss": 1.1068,
+      "step": 12032
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.921279934922447e-05,
+      "loss": 3.9616,
+      "step": 12033
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.935607934496712e-05,
+      "loss": 1.7154,
+      "step": 12034
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.949938144559705e-05,
+      "loss": 4.546,
+      "step": 12035
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.964270535351063e-05,
+      "loss": 2.207,
+      "step": 12036
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.978605077105756e-05,
+      "loss": 1.4561,
+      "step": 12037
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.992941740054424e-05,
+      "loss": 4.2657,
+      "step": 12038
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.007280494423163e-05,
+      "loss": 3.2309,
+      "step": 12039
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.021621310433866e-05,
+      "loss": 1.3026,
+      "step": 12040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.035964158304005e-05,
+      "loss": 1.2447,
+      "step": 12041
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.050309008246973e-05,
+      "loss": 2.3753,
+      "step": 12042
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.06465583047186e-05,
+      "loss": 2.0704,
+      "step": 12043
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.079004595183808e-05,
+      "loss": 0.4558,
+      "step": 12044
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.09335527258378e-05,
+      "loss": 0.7413,
+      "step": 12045
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.10770783286887e-05,
+      "loss": 0.7902,
+      "step": 12046
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.122062246232268e-05,
+      "loss": 3.7061,
+      "step": 12047
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.136418482863204e-05,
+      "loss": 1.4438,
+      "step": 12048
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.150776512947265e-05,
+      "loss": 0.9384,
+      "step": 12049
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.165136306666173e-05,
+      "loss": 0.3347,
+      "step": 12050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.17949783419813e-05,
+      "loss": 3.8251,
+      "step": 12051
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.19386106571759e-05,
+      "loss": 0.1464,
+      "step": 12052
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.208225971395614e-05,
+      "loss": 4.0286,
+      "step": 12053
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.222592521399647e-05,
+      "loss": 3.4589,
+      "step": 12054
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.236960685893854e-05,
+      "loss": 0.6817,
+      "step": 12055
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.251330435038914e-05,
+      "loss": 2.1624,
+      "step": 12056
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.265701738992349e-05,
+      "loss": 2.8499,
+      "step": 12057
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.280074567908313e-05,
+      "loss": 2.5521,
+      "step": 12058
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.294448891937934e-05,
+      "loss": 3.6878,
+      "step": 12059
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.308824681229098e-05,
+      "loss": 1.5132,
+      "step": 12060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.323201905926784e-05,
+      "loss": 0.5347,
+      "step": 12061
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.337580536172851e-05,
+      "loss": 2.9353,
+      "step": 12062
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.351960542106378e-05,
+      "loss": 2.7603,
+      "step": 12063
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.36634189386345e-05,
+      "loss": 4.5047,
+      "step": 12064
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.380724561577496e-05,
+      "loss": 2.1684,
+      "step": 12065
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.395108515379067e-05,
+      "loss": 3.1755,
+      "step": 12066
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.409493725396194e-05,
+      "loss": 0.5371,
+      "step": 12067
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.423880161754146e-05,
+      "loss": 0.8868,
+      "step": 12068
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.438267794575797e-05,
+      "loss": 1.8886,
+      "step": 12069
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.452656593981389e-05,
+      "loss": 3.2102,
+      "step": 12070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.467046530088882e-05,
+      "loss": 1.1702,
+      "step": 12071
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.48143757301374e-05,
+      "loss": 0.1688,
+      "step": 12072
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.495829692869262e-05,
+      "loss": 1.7279,
+      "step": 12073
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.510222859766373e-05,
+      "loss": 1.7143,
+      "step": 12074
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.524617043813966e-05,
+      "loss": 3.3192,
+      "step": 12075
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.539012215118681e-05,
+      "loss": 1.0312,
+      "step": 12076
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.553408343785208e-05,
+      "loss": 1.225,
+      "step": 12077
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.567805399916254e-05,
+      "loss": 2.4489,
+      "step": 12078
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.582203353612493e-05,
+      "loss": 1.0839,
+      "step": 12079
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.596602174972876e-05,
+      "loss": 1.0098,
+      "step": 12080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.611001834094409e-05,
+      "loss": 3.1502,
+      "step": 12081
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.625402301072502e-05,
+      "loss": 0.9951,
+      "step": 12082
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.639803546000743e-05,
+      "loss": 2.4621,
+      "step": 12083
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.654205538971251e-05,
+      "loss": 2.836,
+      "step": 12084
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.668608250074444e-05,
+      "loss": 0.9924,
+      "step": 12085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.683011649399393e-05,
+      "loss": 1.2946,
+      "step": 12086
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.697415707033598e-05,
+      "loss": 0.9477,
+      "step": 12087
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.711820393063335e-05,
+      "loss": 3.0423,
+      "step": 12088
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.72622567757343e-05,
+      "loss": 0.9183,
+      "step": 12089
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.74063153064761e-05,
+      "loss": 3.1055,
+      "step": 12090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.755037922368281e-05,
+      "loss": 1.2153,
+      "step": 12091
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.769444822816867e-05,
+      "loss": 2.5333,
+      "step": 12092
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.783852202073599e-05,
+      "loss": 2.7564,
+      "step": 12093
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.79826003021785e-05,
+      "loss": 2.1268,
+      "step": 12094
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.812668277327927e-05,
+      "loss": 3.3237,
+      "step": 12095
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.8270769134814e-05,
+      "loss": 2.7909,
+      "step": 12096
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.841485908754897e-05,
+      "loss": 2.3184,
+      "step": 12097
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.855895233224434e-05,
+      "loss": 2.3489,
+      "step": 12098
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.870304856965209e-05,
+      "loss": 0.6065,
+      "step": 12099
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.884714750051935e-05,
+      "loss": 1.9243,
+      "step": 12100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.899124882558626e-05,
+      "loss": 3.8327,
+      "step": 12101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.913535224558945e-05,
+      "loss": 0.1326,
+      "step": 12102
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.927945746125966e-05,
+      "loss": 0.7354,
+      "step": 12103
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.942356417332543e-05,
+      "loss": 1.4609,
+      "step": 12104
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.956767208251076e-05,
+      "loss": 1.2727,
+      "step": 12105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.971178088953853e-05,
+      "loss": 0.4736,
+      "step": 12106
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.985589029512839e-05,
+      "loss": 0.518,
+      "step": 12107
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.999999999999976e-05,
+      "loss": 1.2901,
+      "step": 12108
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001001441097048715,
+      "loss": 2.0935,
+      "step": 12109
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010028821911046133,
+      "loss": 1.5905,
+      "step": 12110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001004323279174891,
+      "loss": 1.0588,
+      "step": 12111
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010057643582667444,
+      "loss": 1.4629,
+      "step": 12112
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010072054253874021,
+      "loss": 1.8013,
+      "step": 12113
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010086464775441044,
+      "loss": 3.5181,
+      "step": 12114
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001010087511744136,
+      "loss": 2.2175,
+      "step": 12115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010115285249948052,
+      "loss": 2.1235,
+      "step": 12116
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010129695143034778,
+      "loss": 1.2973,
+      "step": 12117
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010144104766775553,
+      "loss": 0.8117,
+      "step": 12118
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010158514091245091,
+      "loss": 1.4556,
+      "step": 12119
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010172923086518587,
+      "loss": 1.0777,
+      "step": 12120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010187331722672062,
+      "loss": 6.5322,
+      "step": 12121
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010201739969782139,
+      "loss": 2.9624,
+      "step": 12122
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001021614779792639,
+      "loss": 1.4022,
+      "step": 12123
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001023055517718312,
+      "loss": 1.4349,
+      "step": 12124
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010244962077631708,
+      "loss": 1.8004,
+      "step": 12125
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010259368469352377,
+      "loss": 0.8154,
+      "step": 12126
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010273774322426557,
+      "loss": 0.768,
+      "step": 12127
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010288179606936652,
+      "loss": 1.8245,
+      "step": 12128
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010302584292966389,
+      "loss": 2.6882,
+      "step": 12129
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010316988350600593,
+      "loss": 0.5216,
+      "step": 12130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010331391749925545,
+      "loss": 2.4615,
+      "step": 12131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010345794461028737,
+      "loss": 3.6695,
+      "step": 12132
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010360196453999245,
+      "loss": 2.0121,
+      "step": 12133
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010374597698927486,
+      "loss": 0.7293,
+      "step": 12134
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001038899816590558,
+      "loss": 1.4402,
+      "step": 12135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010403397825027112,
+      "loss": 0.2588,
+      "step": 12136
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010417796646387494,
+      "loss": 0.1687,
+      "step": 12137
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010432194600083734,
+      "loss": 2.1063,
+      "step": 12138
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010446591656214779,
+      "loss": 1.8074,
+      "step": 12139
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010460987784881306,
+      "loss": 1.5486,
+      "step": 12140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001047538295618602,
+      "loss": 4.1028,
+      "step": 12141
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010489777140233614,
+      "loss": 1.1222,
+      "step": 12142
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010504170307130727,
+      "loss": 1.2972,
+      "step": 12143
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010518562426986249,
+      "loss": 0.918,
+      "step": 12144
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010532953469911105,
+      "loss": 0.2359,
+      "step": 12145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000105473434060186,
+      "loss": 1.7624,
+      "step": 12146
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001056173220542419,
+      "loss": 0.2194,
+      "step": 12147
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010576119838245842,
+      "loss": 1.6948,
+      "step": 12148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010590506274603794,
+      "loss": 0.8996,
+      "step": 12149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001060489148462092,
+      "loss": 1.0284,
+      "step": 12150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010619275438422493,
+      "loss": 2.1841,
+      "step": 12151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010633658106136537,
+      "loss": 1.6798,
+      "step": 12152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010648039457893608,
+      "loss": 3.0572,
+      "step": 12153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010662419463827138,
+      "loss": 1.4057,
+      "step": 12154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010676798094073203,
+      "loss": 1.1474,
+      "step": 12155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010691175318770889,
+      "loss": 0.8413,
+      "step": 12156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010705551108062053,
+      "loss": 0.7889,
+      "step": 12157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010719925432091676,
+      "loss": 0.6417,
+      "step": 12158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010734298261007638,
+      "loss": 0.3255,
+      "step": 12159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010748669564961073,
+      "loss": 2.1606,
+      "step": 12160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010763039314106132,
+      "loss": 0.5701,
+      "step": 12161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010777407478600341,
+      "loss": 0.7124,
+      "step": 12162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010791774028604373,
+      "loss": 1.0952,
+      "step": 12163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000108061389342824,
+      "loss": 2.7154,
+      "step": 12164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010820502165801858,
+      "loss": 2.2271,
+      "step": 12165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010834863693333814,
+      "loss": 0.7414,
+      "step": 12166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010849223487052722,
+      "loss": 1.0183,
+      "step": 12167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010863581517136785,
+      "loss": 5.974,
+      "step": 12168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001087793775376772,
+      "loss": 1.5958,
+      "step": 12169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010892292167131117,
+      "loss": 1.9097,
+      "step": 12170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010906644727416206,
+      "loss": 2.9244,
+      "step": 12171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010920995404816179,
+      "loss": 0.8372,
+      "step": 12172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010935344169528126,
+      "loss": 4.8338,
+      "step": 12173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010949690991753014,
+      "loss": 5.2843,
+      "step": 12174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010964035841695981,
+      "loss": 2.7313,
+      "step": 12175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001097837868956612,
+      "loss": 0.4891,
+      "step": 12176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010992719505576826,
+      "loss": 1.2602,
+      "step": 12177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011007058259945564,
+      "loss": 2.9274,
+      "step": 12178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011021394922894233,
+      "loss": 1.9321,
+      "step": 12179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011035729464648923,
+      "loss": 4.7776,
+      "step": 12180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011050061855440281,
+      "loss": 0.3169,
+      "step": 12181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011064392065503275,
+      "loss": 2.9209,
+      "step": 12182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011078720065077541,
+      "loss": 1.8349,
+      "step": 12183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011093045824407167,
+      "loss": 1.6753,
+      "step": 12184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011107369313741034,
+      "loss": 3.4755,
+      "step": 12185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011121690503332595,
+      "loss": 2.6269,
+      "step": 12186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011136009363440225,
+      "loss": 3.1482,
+      "step": 12187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001115032586432699,
+      "loss": 2.6832,
+      "step": 12188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011164639976260996,
+      "loss": 0.1674,
+      "step": 12189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011178951669515174,
+      "loss": 1.345,
+      "step": 12190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001119326091436762,
+      "loss": 1.7279,
+      "step": 12191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011207567681101365,
+      "loss": 1.026,
+      "step": 12192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011221871940004737,
+      "loss": 2.2043,
+      "step": 12193
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011236173661371126,
+      "loss": 1.1944,
+      "step": 12194
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011250472815499335,
+      "loss": 0.1191,
+      "step": 12195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011264769372693356,
+      "loss": 4.9534,
+      "step": 12196
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011279063303262715,
+      "loss": 1.2977,
+      "step": 12197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011293354577522255,
+      "loss": 2.2696,
+      "step": 12198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011307643165792472,
+      "loss": 3.1078,
+      "step": 12199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011321929038399302,
+      "loss": 1.0592,
+      "step": 12200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011336212165674462,
+      "loss": 2.7609,
+      "step": 12201
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011350492517955229,
+      "loss": 1.6365,
+      "step": 12202
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011364770065584747,
+      "loss": 0.6806,
+      "step": 12203
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00011379044778911986,
+      "loss": 1.4003,
+      "step": 12204
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000113933166282917,
+      "loss": 2.3744,
+      "step": 12205
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011407585584084723,
+      "loss": 0.3327,
+      "step": 12206
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011421851616657766,
+      "loss": 0.7832,
+      "step": 12207
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011436114696383746,
+      "loss": 0.6819,
+      "step": 12208
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011450374793641576,
+      "loss": 3.1517,
+      "step": 12209
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011464631878816499,
+      "loss": 1.6369,
+      "step": 12210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011478885922299875,
+      "loss": 4.8698,
+      "step": 12211
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011493136894489525,
+      "loss": 2.3021,
+      "step": 12212
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011507384765789502,
+      "loss": 5.1727,
+      "step": 12213
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001152162950661044,
+      "loss": 2.9099,
+      "step": 12214
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011535871087369338,
+      "loss": 0.3482,
+      "step": 12215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011550109478489894,
+      "loss": 1.1303,
+      "step": 12216
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011564344650402294,
+      "loss": 4.0888,
+      "step": 12217
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011578576573543544,
+      "loss": 0.1415,
+      "step": 12218
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011592805218357261,
+      "loss": 0.9524,
+      "step": 12219
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011607030555294009,
+      "loss": 0.8072,
+      "step": 12220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011621252554811081,
+      "loss": 3.9047,
+      "step": 12221
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011635471187372844,
+      "loss": 3.6242,
+      "step": 12222
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011649686423450517,
+      "loss": 0.8916,
+      "step": 12223
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011663898233522505,
+      "loss": 2.8081,
+      "step": 12224
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011678106588074198,
+      "loss": 2.812,
+      "step": 12225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000116923114575983,
+      "loss": 0.9296,
+      "step": 12226
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011706512812594606,
+      "loss": 2.0327,
+      "step": 12227
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011720710623570356,
+      "loss": 0.148,
+      "step": 12228
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011734904861040012,
+      "loss": 2.7728,
+      "step": 12229
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011749095495525593,
+      "loss": 2.8378,
+      "step": 12230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001176328249755646,
+      "loss": 2.1403,
+      "step": 12231
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011777465837669659,
+      "loss": 3.555,
+      "step": 12232
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011791645486409703,
+      "loss": 1.3653,
+      "step": 12233
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011805821414328911,
+      "loss": 3.4875,
+      "step": 12234
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011819993591987185,
+      "loss": 0.408,
+      "step": 12235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011834161989952322,
+      "loss": 1.1377,
+      "step": 12236
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001184832657879997,
+      "loss": 1.6904,
+      "step": 12237
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011862487329113585,
+      "loss": 1.3908,
+      "step": 12238
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001187664421148473,
+      "loss": 1.4047,
+      "step": 12239
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011890797196512864,
+      "loss": 2.3175,
+      "step": 12240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011904946254805676,
+      "loss": 0.5088,
+      "step": 12241
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011919091356978875,
+      "loss": 2.4724,
+      "step": 12242
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011933232473656524,
+      "loss": 1.0843,
+      "step": 12243
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011947369575470819,
+      "loss": 4.1972,
+      "step": 12244
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011961502633062438,
+      "loss": 1.5797,
+      "step": 12245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011975631617080321,
+      "loss": 0.9637,
+      "step": 12246
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00011989756498182002,
+      "loss": 0.8878,
+      "step": 12247
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012003877247033393,
+      "loss": 2.3937,
+      "step": 12248
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012017993834309141,
+      "loss": 0.1975,
+      "step": 12249
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001203210623069238,
+      "loss": 3.131,
+      "step": 12250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012046214406875096,
+      "loss": 0.8161,
+      "step": 12251
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000120603183335579,
+      "loss": 3.1303,
+      "step": 12252
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012074417981450365,
+      "loss": 0.5814,
+      "step": 12253
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012088513321270811,
+      "loss": 0.435,
+      "step": 12254
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012102604323746645,
+      "loss": 2.0027,
+      "step": 12255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012116690959614141,
+      "loss": 2.8144,
+      "step": 12256
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012130773199618786,
+      "loss": 1.3896,
+      "step": 12257
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012144851014515044,
+      "loss": 4.042,
+      "step": 12258
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012158924375066725,
+      "loss": 1.5167,
+      "step": 12259
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012172993252046738,
+      "loss": 2.7422,
+      "step": 12260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012187057616237447,
+      "loss": 4.3995,
+      "step": 12261
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012201117438430449,
+      "loss": 3.0475,
+      "step": 12262
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012215172689426912,
+      "loss": 2.1656,
+      "step": 12263
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012229223340037357,
+      "loss": 3.407,
+      "step": 12264
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012243269361082002,
+      "loss": 1.8495,
+      "step": 12265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012257310723390536,
+      "loss": 1.8195,
+      "step": 12266
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001227134739780243,
+      "loss": 0.9987,
+      "step": 12267
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012285379355166887,
+      "loss": 1.4315,
+      "step": 12268
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012299406566342809,
+      "loss": 0.2828,
+      "step": 12269
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012313429002199086,
+      "loss": 1.0414,
+      "step": 12270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012327446633614392,
+      "loss": 2.9565,
+      "step": 12271
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012341459431477512,
+      "loss": 1.8889,
+      "step": 12272
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001235546736668714,
+      "loss": 2.9533,
+      "step": 12273
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012369470410152192,
+      "loss": 2.0711,
+      "step": 12274
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001238346853279162,
+      "loss": 2.1574,
+      "step": 12275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012397461705534728,
+      "loss": 1.5707,
+      "step": 12276
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012411449899320956,
+      "loss": 1.1046,
+      "step": 12277
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012425433085100224,
+      "loss": 3.5055,
+      "step": 12278
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012439411233832725,
+      "loss": 2.1373,
+      "step": 12279
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001245338431648924,
+      "loss": 2.4172,
+      "step": 12280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012467352304050935,
+      "loss": 2.3456,
+      "step": 12281
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012481315167509695,
+      "loss": 0.2126,
+      "step": 12282
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012495272877867912,
+      "loss": 2.4265,
+      "step": 12283
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012509225406138813,
+      "loss": 3.4809,
+      "step": 12284
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012523172723346256,
+      "loss": 1.3058,
+      "step": 12285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012537114800525051,
+      "loss": 1.9015,
+      "step": 12286
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001255105160872076,
+      "loss": 2.4057,
+      "step": 12287
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012564983118990014,
+      "loss": 1.5811,
+      "step": 12288
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012578909302400323,
+      "loss": 3.262,
+      "step": 12289
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012592830130030388,
+      "loss": 2.9771,
+      "step": 12290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012606745572969898,
+      "loss": 1.9885,
+      "step": 12291
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012620655602319858,
+      "loss": 3.0087,
+      "step": 12292
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012634560189192387,
+      "loss": 0.4589,
+      "step": 12293
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012648459304711037,
+      "loss": 4.846,
+      "step": 12294
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012662352920010585,
+      "loss": 3.4707,
+      "step": 12295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012676241006237375,
+      "loss": 4.1245,
+      "step": 12296
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001269012353454909,
+      "loss": 1.7108,
+      "step": 12297
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012704000476115057,
+      "loss": 1.0774,
+      "step": 12298
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012717871802116212,
+      "loss": 2.2416,
+      "step": 12299
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012731737483745044,
+      "loss": 3.6704,
+      "step": 12300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001274559749220591,
+      "loss": 2.0052,
+      "step": 12301
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012759451798714796,
+      "loss": 2.768,
+      "step": 12302
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012773300374499684,
+      "loss": 2.4281,
+      "step": 12303
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001278714319080031,
+      "loss": 1.2997,
+      "step": 12304
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012800980218868508,
+      "loss": 2.5867,
+      "step": 12305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012814811429968004,
+      "loss": 2.5065,
+      "step": 12306
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001282863679537473,
+      "loss": 4.1402,
+      "step": 12307
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001284245628637663,
+      "loss": 1.0487,
+      "step": 12308
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012856269874273986,
+      "loss": 1.2339,
+      "step": 12309
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012870077530379188,
+      "loss": 2.2593,
+      "step": 12310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000128838792260171,
+      "loss": 2.7749,
+      "step": 12311
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001289767493252482,
+      "loss": 3.5383,
+      "step": 12312
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012911464621252012,
+      "loss": 4.8025,
+      "step": 12313
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012925248263560718,
+      "loss": 1.6654,
+      "step": 12314
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001293902583082566,
+      "loss": 3.4227,
+      "step": 12315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012952797294434046,
+      "loss": 2.7763,
+      "step": 12316
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012966562625785897,
+      "loss": 0.402,
+      "step": 12317
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012980321796293826,
+      "loss": 0.5881,
+      "step": 12318
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00012994074777383387,
+      "loss": 2.016,
+      "step": 12319
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013007821540492843,
+      "loss": 1.6194,
+      "step": 12320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013021562057073508,
+      "loss": 2.7168,
+      "step": 12321
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001303529629858954,
+      "loss": 2.4345,
+      "step": 12322
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013049024236518254,
+      "loss": 0.2661,
+      "step": 12323
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013062745842349927,
+      "loss": 0.7611,
+      "step": 12324
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001307646108758812,
+      "loss": 0.4956,
+      "step": 12325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013090169943749468,
+      "loss": 2.1386,
+      "step": 12326
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013103872382364012,
+      "loss": 0.5164,
+      "step": 12327
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013117568374974983,
+      "loss": 2.3182,
+      "step": 12328
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013131257893139096,
+      "loss": 1.6436,
+      "step": 12329
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013144940908426525,
+      "loss": 1.3278,
+      "step": 12330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013158617392420834,
+      "loss": 1.6841,
+      "step": 12331
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001317228731671929,
+      "loss": 2.6798,
+      "step": 12332
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013185950652932652,
+      "loss": 2.4248,
+      "step": 12333
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013199607372685494,
+      "loss": 0.5536,
+      "step": 12334
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013213257447615992,
+      "loss": 1.2447,
+      "step": 12335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001322690084937626,
+      "loss": 3.1221,
+      "step": 12336
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013240537549632134,
+      "loss": 2.5954,
+      "step": 12337
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013254167520063509,
+      "loss": 2.4872,
+      "step": 12338
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013267790732364112,
+      "loss": 0.7176,
+      "step": 12339
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013281407158241844,
+      "loss": 1.3901,
+      "step": 12340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001329501676941856,
+      "loss": 1.0983,
+      "step": 12341
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013308619537630414,
+      "loss": 0.3406,
+      "step": 12342
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013322215434627633,
+      "loss": 4.7702,
+      "step": 12343
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013335804432174832,
+      "loss": 2.3054,
+      "step": 12344
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013349386502050845,
+      "loss": 1.4813,
+      "step": 12345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013362961616049008,
+      "loss": 0.3388,
+      "step": 12346
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001337652974597698,
+      "loss": 2.4433,
+      "step": 12347
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013390090863657053,
+      "loss": 2.2576,
+      "step": 12348
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013403644940925948,
+      "loss": 0.4363,
+      "step": 12349
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013417191949635144,
+      "loss": 0.6547,
+      "step": 12350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013430731861650665,
+      "loss": 1.8033,
+      "step": 12351
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000134442646488534,
+      "loss": 2.8478,
+      "step": 12352
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013457790283138917,
+      "loss": 4.3521,
+      "step": 12353
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013471308736417756,
+      "loss": 2.4712,
+      "step": 12354
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001348481998061524,
+      "loss": 0.6056,
+      "step": 12355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001349832398767181,
+      "loss": 1.8548,
+      "step": 12356
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013511820729542786,
+      "loss": 1.1511,
+      "step": 12357
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001352531017819872,
+      "loss": 2.657,
+      "step": 12358
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001353879230562517,
+      "loss": 2.2707,
+      "step": 12359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013552267083823038,
+      "loss": 1.7252,
+      "step": 12360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013565734484808357,
+      "loss": 1.7777,
+      "step": 12361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013579194480612568,
+      "loss": 1.4146,
+      "step": 12362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001359264704328251,
+      "loss": 1.8272,
+      "step": 12363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013606092144880338,
+      "loss": 0.6456,
+      "step": 12364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013619529757483853,
+      "loss": 0.836,
+      "step": 12365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013632959853186264,
+      "loss": 3.1524,
+      "step": 12366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001364638240409653,
+      "loss": 0.6697,
+      "step": 12367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013659797382339144,
+      "loss": 0.2981,
+      "step": 12368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013673204760054458,
+      "loss": 0.1684,
+      "step": 12369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013686604509398473,
+      "loss": 2.2175,
+      "step": 12370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001369999660254317,
+      "loss": 2.5667,
+      "step": 12371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013713381011676298,
+      "loss": 4.5369,
+      "step": 12372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001372675770900169,
+      "loss": 1.2255,
+      "step": 12373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013740126666739072,
+      "loss": 1.2462,
+      "step": 12374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013753487857124362,
+      "loss": 0.6211,
+      "step": 12375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001376684125240949,
+      "loss": 3.7155,
+      "step": 12376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013780186824862697,
+      "loss": 0.3438,
+      "step": 12377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013793524546768345,
+      "loss": 0.3295,
+      "step": 12378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001380685439042723,
+      "loss": 5.3981,
+      "step": 12379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001382017632815637,
+      "loss": 2.3358,
+      "step": 12380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013833490332289348,
+      "loss": 1.033,
+      "step": 12381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013846796375176074,
+      "loss": 4.7122,
+      "step": 12382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013860094429183142,
+      "loss": 1.3339,
+      "step": 12383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001387338446669359,
+      "loss": 2.1632,
+      "step": 12384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013886666460107244,
+      "loss": 0.7537,
+      "step": 12385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000138999403818405,
+      "loss": 2.4692,
+      "step": 12386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013913206204326653,
+      "loss": 2.136,
+      "step": 12387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00013926463900015682,
+      "loss": 0.4029,
+      "step": 12388
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013939713441374577,
+      "loss": 0.8917,
+      "step": 12389
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013952954800887133,
+      "loss": 0.7568,
+      "step": 12390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013966187951054267,
+      "loss": 2.0931,
+      "step": 12391
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013979412864393806,
+      "loss": 1.791,
+      "step": 12392
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00013992629513440795,
+      "loss": 2.2971,
+      "step": 12393
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014005837870747435,
+      "loss": 2.5059,
+      "step": 12394
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014019037908883047,
+      "loss": 3.3784,
+      "step": 12395
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001403222960043433,
+      "loss": 2.1973,
+      "step": 12396
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014045412918005314,
+      "loss": 1.0478,
+      "step": 12397
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014058587834217355,
+      "loss": 2.1369,
+      "step": 12398
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014071754321709262,
+      "loss": 5.2631,
+      "step": 12399
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014084912353137444,
+      "loss": 4.3097,
+      "step": 12400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001409806190117573,
+      "loss": 4.4483,
+      "step": 12401
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014111202938515714,
+      "loss": 0.5566,
+      "step": 12402
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014124335437866524,
+      "loss": 0.765,
+      "step": 12403
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014137459371955154,
+      "loss": 3.265,
+      "step": 12404
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001415057471352625,
+      "loss": 0.8227,
+      "step": 12405
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014163681435342443,
+      "loss": 1.7217,
+      "step": 12406
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014176779510184125,
+      "loss": 1.7707,
+      "step": 12407
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001418986891084978,
+      "loss": 4.0112,
+      "step": 12408
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001420294961015578,
+      "loss": 2.9331,
+      "step": 12409
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014216021580936686,
+      "loss": 1.4845,
+      "step": 12410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014229084796045068,
+      "loss": 0.8898,
+      "step": 12411
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014242139228351809,
+      "loss": 2.0091,
+      "step": 12412
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014255184850745889,
+      "loss": 3.521,
+      "step": 12413
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001426822163613473,
+      "loss": 1.6668,
+      "step": 12414
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014281249557443967,
+      "loss": 1.5666,
+      "step": 12415
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014294268587617775,
+      "loss": 1.4408,
+      "step": 12416
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001430727869961867,
+      "loss": 1.6176,
+      "step": 12417
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014320279866427807,
+      "loss": 1.6523,
+      "step": 12418
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014333272061044797,
+      "loss": 2.5933,
+      "step": 12419
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001434625525648801,
+      "loss": 2.3091,
+      "step": 12420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014359229425794383,
+      "loss": 1.3487,
+      "step": 12421
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014372194542019715,
+      "loss": 2.0289,
+      "step": 12422
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014385150578238488,
+      "loss": 4.1083,
+      "step": 12423
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014398097507544127,
+      "loss": 2.3645,
+      "step": 12424
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014411035303048985,
+      "loss": 1.589,
+      "step": 12425
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014423963937884273,
+      "loss": 0.1837,
+      "step": 12426
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014436883385200364,
+      "loss": 2.1243,
+      "step": 12427
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014449793618166574,
+      "loss": 0.5617,
+      "step": 12428
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014462694609971492,
+      "loss": 0.1739,
+      "step": 12429
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014475586333822767,
+      "loss": 1.7893,
+      "step": 12430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014488468762947427,
+      "loss": 1.2948,
+      "step": 12431
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014501341870591666,
+      "loss": 1.0942,
+      "step": 12432
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001451420563002117,
+      "loss": 2.3581,
+      "step": 12433
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014527060014520916,
+      "loss": 4.8635,
+      "step": 12434
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001453990499739547,
+      "loss": 2.6398,
+      "step": 12435
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000145527405519688,
+      "loss": 3.3812,
+      "step": 12436
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014565566651584578,
+      "loss": 1.1653,
+      "step": 12437
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014578383269605991,
+      "loss": 2.1216,
+      "step": 12438
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014591190379416037,
+      "loss": 0.8904,
+      "step": 12439
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014603987954417338,
+      "loss": 0.3368,
+      "step": 12440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014616775968032442,
+      "loss": 3.1978,
+      "step": 12441
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014629554393703624,
+      "loss": 0.9587,
+      "step": 12442
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001464232320489321,
+      "loss": 3.1949,
+      "step": 12443
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014655082375083348,
+      "loss": 0.3755,
+      "step": 12444
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001466783187777635,
+      "loss": 1.0442,
+      "step": 12445
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014680571686494473,
+      "loss": 1.6299,
+      "step": 12446
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001469330177478023,
+      "loss": 2.2213,
+      "step": 12447
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000147060221161962,
+      "loss": 1.3352,
+      "step": 12448
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014718732684325324,
+      "loss": 2.2737,
+      "step": 12449
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014731433452770716,
+      "loss": 1.7248,
+      "step": 12450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014744124395155972,
+      "loss": 0.2005,
+      "step": 12451
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014756805485124962,
+      "loss": 0.2022,
+      "step": 12452
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014769476696342148,
+      "loss": 1.0328,
+      "step": 12453
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001478213800249238,
+      "loss": 1.7626,
+      "step": 12454
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014794789377281173,
+      "loss": 2.7092,
+      "step": 12455
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014807430794434664,
+      "loss": 2.1909,
+      "step": 12456
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001482006222769958,
+      "loss": 2.657,
+      "step": 12457
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014832683650843506,
+      "loss": 3.8619,
+      "step": 12458
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001484529503765469,
+      "loss": 0.6244,
+      "step": 12459
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014857896361942317,
+      "loss": 0.9344,
+      "step": 12460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001487048759753647,
+      "loss": 1.8174,
+      "step": 12461
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014883068718288118,
+      "loss": 2.5481,
+      "step": 12462
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014895639698069243,
+      "loss": 2.5353,
+      "step": 12463
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014908200510772978,
+      "loss": 3.5083,
+      "step": 12464
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001492075113031344,
+      "loss": 4.2222,
+      "step": 12465
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014933291530626046,
+      "loss": 3.2501,
+      "step": 12466
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014945821685667313,
+      "loss": 2.4766,
+      "step": 12467
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001495834156941515,
+      "loss": 1.7729,
+      "step": 12468
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014970851155868685,
+      "loss": 3.3419,
+      "step": 12469
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014983350419048546,
+      "loss": 0.5987,
+      "step": 12470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00014995839332996677,
+      "loss": 0.2122,
+      "step": 12471
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015008317871776646,
+      "loss": 1.841,
+      "step": 12472
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015020786009473433,
+      "loss": 1.2233,
+      "step": 12473
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015033243720193753,
+      "loss": 0.7563,
+      "step": 12474
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015045690978065843,
+      "loss": 0.6076,
+      "step": 12475
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015058127757239777,
+      "loss": 2.5539,
+      "step": 12476
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015070554031887263,
+      "loss": 0.2399,
+      "step": 12477
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015082969776201955,
+      "loss": 4.2602,
+      "step": 12478
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015095374964399245,
+      "loss": 2.6099,
+      "step": 12479
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015107769570716574,
+      "loss": 0.3339,
+      "step": 12480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015120153569413237,
+      "loss": 0.8383,
+      "step": 12481
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001513252693477068,
+      "loss": 0.5172,
+      "step": 12482
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015144889641092307,
+      "loss": 0.982,
+      "step": 12483
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015157241662703785,
+      "loss": 1.4599,
+      "step": 12484
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015169582973952847,
+      "loss": 1.2259,
+      "step": 12485
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015181913549209593,
+      "loss": 0.5688,
+      "step": 12486
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015194233362866288,
+      "loss": 0.4346,
+      "step": 12487
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001520654238933765,
+      "loss": 3.8017,
+      "step": 12488
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001521884060306079,
+      "loss": 0.9013,
+      "step": 12489
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001523112797849519,
+      "loss": 2.0724,
+      "step": 12490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015243404490122928,
+      "loss": 0.3659,
+      "step": 12491
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001525567011244864,
+      "loss": 1.1772,
+      "step": 12492
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015267924819999527,
+      "loss": 2.1258,
+      "step": 12493
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015280168587325445,
+      "loss": 1.6,
+      "step": 12494
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001529240138899907,
+      "loss": 2.734,
+      "step": 12495
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015304623199615723,
+      "loss": 2.052,
+      "step": 12496
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015316833993793676,
+      "loss": 2.2098,
+      "step": 12497
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015329033746173961,
+      "loss": 2.091,
+      "step": 12498
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015341222431420657,
+      "loss": 4.6205,
+      "step": 12499
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001535340002422071,
+      "loss": 2.3488,
+      "step": 12500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015365566499284214,
+      "loss": 1.3599,
+      "step": 12501
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015377721831344245,
+      "loss": 0.7012,
+      "step": 12502
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001538986599515713,
+      "loss": 0.7271,
+      "step": 12503
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001540199896550228,
+      "loss": 1.5892,
+      "step": 12504
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001541412071718246,
+      "loss": 0.8089,
+      "step": 12505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001542623122502362,
+      "loss": 3.0027,
+      "step": 12506
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015438330463875185,
+      "loss": 2.4318,
+      "step": 12507
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001545041840860985,
+      "loss": 0.67,
+      "step": 12508
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015462495034123901,
+      "loss": 1.5957,
+      "step": 12509
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015474560315337,
+      "loss": 2.9906,
+      "step": 12510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001548661422719249,
+      "loss": 2.6072,
+      "step": 12511
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015498656744657212,
+      "loss": 2.209,
+      "step": 12512
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015510687842721784,
+      "loss": 1.3009,
+      "step": 12513
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001552270749640042,
+      "loss": 3.0564,
+      "step": 12514
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015534715680731227,
+      "loss": 2.6545,
+      "step": 12515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015546712370776003,
+      "loss": 3.8127,
+      "step": 12516
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001555869754162054,
+      "loss": 4.574,
+      "step": 12517
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015570671168374433,
+      "loss": 0.6912,
+      "step": 12518
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015582633226171347,
+      "loss": 0.3035,
+      "step": 12519
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001559458369016896,
+      "loss": 3.3973,
+      "step": 12520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015606522535548952,
+      "loss": 5.1939,
+      "step": 12521
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001561844973751721,
+      "loss": 3.3632,
+      "step": 12522
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001563036527130381,
+      "loss": 1.8138,
+      "step": 12523
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015642269112162965,
+      "loss": 5.6637,
+      "step": 12524
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015654161235373294,
+      "loss": 1.9267,
+      "step": 12525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001566604161623766,
+      "loss": 0.9472,
+      "step": 12526
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001567791023008331,
+      "loss": 0.7728,
+      "step": 12527
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015689767052262026,
+      "loss": 2.0821,
+      "step": 12528
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015701612058149946,
+      "loss": 0.4347,
+      "step": 12529
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001571344522314788,
+      "loss": 1.3794,
+      "step": 12530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015725266522681098,
+      "loss": 1.7573,
+      "step": 12531
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001573707593219964,
+      "loss": 0.909,
+      "step": 12532
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015748873427178113,
+      "loss": 2.7265,
+      "step": 12533
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015760658983115996,
+      "loss": 2.74,
+      "step": 12534
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015772432575537435,
+      "loss": 0.6143,
+      "step": 12535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015784194179991538,
+      "loss": 1.4349,
+      "step": 12536
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015795943772052202,
+      "loss": 3.2846,
+      "step": 12537
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015807681327318376,
+      "loss": 1.9426,
+      "step": 12538
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000158194068214139,
+      "loss": 2.2512,
+      "step": 12539
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001583112022998777,
+      "loss": 1.2823,
+      "step": 12540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015842821528713969,
+      "loss": 2.8528,
+      "step": 12541
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001585451069329175,
+      "loss": 1.7096,
+      "step": 12542
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015866187699445442,
+      "loss": 1.0155,
+      "step": 12543
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001587785252292474,
+      "loss": 0.3509,
+      "step": 12544
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001588950513950453,
+      "loss": 1.8006,
+      "step": 12545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015901145524985162,
+      "loss": 3.0095,
+      "step": 12546
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015912773655192265,
+      "loss": 1.4634,
+      "step": 12547
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015924389505977046,
+      "loss": 2.5654,
+      "step": 12548
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001593599305321609,
+      "loss": 3.0108,
+      "step": 12549
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015947584272811623,
+      "loss": 0.6969,
+      "step": 12550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001595916314069147,
+      "loss": 0.782,
+      "step": 12551
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015970729632809025,
+      "loss": 3.8714,
+      "step": 12552
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00015982283725143496,
+      "loss": 2.0546,
+      "step": 12553
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001599382539369973,
+      "loss": 1.7908,
+      "step": 12554
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001600535461450846,
+      "loss": 1.7961,
+      "step": 12555
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016016871363626263,
+      "loss": 5.2396,
+      "step": 12556
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016028375617135573,
+      "loss": 0.3071,
+      "step": 12557
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016039867351144764,
+      "loss": 1.0757,
+      "step": 12558
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016051346541788298,
+      "loss": 2.2715,
+      "step": 12559
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001606281316522658,
+      "loss": 0.9548,
+      "step": 12560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016074267197646215,
+      "loss": 2.8228,
+      "step": 12561
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016085708615259852,
+      "loss": 0.4961,
+      "step": 12562
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016097137394306446,
+      "loss": 1.4795,
+      "step": 12563
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016108553511051094,
+      "loss": 2.8675,
+      "step": 12564
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016119956941785296,
+      "loss": 2.7264,
+      "step": 12565
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016131347662826784,
+      "loss": 2.5263,
+      "step": 12566
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016142725650519798,
+      "loss": 0.4628,
+      "step": 12567
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001615409088123492,
+      "loss": 4.4297,
+      "step": 12568
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001616544333136932,
+      "loss": 0.3484,
+      "step": 12569
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016176782977346616,
+      "loss": 2.2759,
+      "step": 12570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00016188109795617117,
+      "loss": 1.3207,
+      "step": 12571
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001619942376265766,
+      "loss": 1.8834,
+      "step": 12572
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016210724854971896,
+      "loss": 1.6765,
+      "step": 12573
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001622201304909008,
+      "loss": 0.9897,
+      "step": 12574
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016233288321569386,
+      "loss": 1.4799,
+      "step": 12575
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000162445506489937,
+      "loss": 3.1336,
+      "step": 12576
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016255800007973912,
+      "loss": 2.8647,
+      "step": 12577
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016267036375147722,
+      "loss": 2.8445,
+      "step": 12578
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016278259727179928,
+      "loss": 1.3667,
+      "step": 12579
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016289470040762246,
+      "loss": 1.8424,
+      "step": 12580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016300667292613584,
+      "loss": 2.9568,
+      "step": 12581
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001631185145947986,
+      "loss": 2.9447,
+      "step": 12582
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016323022518134248,
+      "loss": 2.815,
+      "step": 12583
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001633418044537715,
+      "loss": 0.4475,
+      "step": 12584
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016345325218036162,
+      "loss": 1.4418,
+      "step": 12585
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001635645681296626,
+      "loss": 0.3505,
+      "step": 12586
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001636757520704981,
+      "loss": 2.9517,
+      "step": 12587
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016378680377196497,
+      "loss": 3.6714,
+      "step": 12588
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016389772300343585,
+      "loss": 0.488,
+      "step": 12589
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016400850953455763,
+      "loss": 1.5002,
+      "step": 12590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001641191631352528,
+      "loss": 1.9009,
+      "step": 12591
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016422968357572074,
+      "loss": 0.7507,
+      "step": 12592
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001643400706264363,
+      "loss": 1.2088,
+      "step": 12593
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001644503240581524,
+      "loss": 0.8152,
+      "step": 12594
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001645604436418984,
+      "loss": 2.3309,
+      "step": 12595
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001646704291489827,
+      "loss": 1.8006,
+      "step": 12596
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000164780280350991,
+      "loss": 3.204,
+      "step": 12597
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001648899970197891,
+      "loss": 3.196,
+      "step": 12598
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016499957892752109,
+      "loss": 2.4931,
+      "step": 12599
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000165109025846612,
+      "loss": 1.566,
+      "step": 12600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016521833754976607,
+      "loss": 1.9241,
+      "step": 12601
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001653275138099695,
+      "loss": 1.2084,
+      "step": 12602
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001654365544004887,
+      "loss": 1.0047,
+      "step": 12603
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001655454590948728,
+      "loss": 2.398,
+      "step": 12604
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016565422766695224,
+      "loss": 0.5768,
+      "step": 12605
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016576285989084108,
+      "loss": 3.0189,
+      "step": 12606
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016587135554093556,
+      "loss": 2.2831,
+      "step": 12607
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016597971439191657,
+      "loss": 1.6397,
+      "step": 12608
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016608793621874807,
+      "loss": 0.6167,
+      "step": 12609
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016619602079667964,
+      "loss": 1.6267,
+      "step": 12610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016630396790124485,
+      "loss": 0.2806,
+      "step": 12611
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016641177730826378,
+      "loss": 0.9081,
+      "step": 12612
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016651944879384148,
+      "loss": 1.3399,
+      "step": 12613
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016662698213437013,
+      "loss": 3.1748,
+      "step": 12614
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016673437710652887,
+      "loss": 0.2901,
+      "step": 12615
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016684163348728347,
+      "loss": 2.4831,
+      "step": 12616
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016694875105388818,
+      "loss": 2.3482,
+      "step": 12617
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016705572958388562,
+      "loss": 4.7546,
+      "step": 12618
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001671625688551063,
+      "loss": 1.8254,
+      "step": 12619
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016726926864567108,
+      "loss": 2.8507,
+      "step": 12620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016737582873398963,
+      "loss": 1.2376,
+      "step": 12621
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016748224889876176,
+      "loss": 1.1201,
+      "step": 12622
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016758852891897867,
+      "loss": 2.8289,
+      "step": 12623
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016769466857392156,
+      "loss": 3.4608,
+      "step": 12624
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016780066764316418,
+      "loss": 4.2139,
+      "step": 12625
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016790652590657117,
+      "loss": 1.3576,
+      "step": 12626
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016801224314430066,
+      "loss": 2.8242,
+      "step": 12627
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016811781913680262,
+      "loss": 1.6291,
+      "step": 12628
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016822325366482139,
+      "loss": 3.5539,
+      "step": 12629
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016832854650939403,
+      "loss": 1.728,
+      "step": 12630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016843369745185292,
+      "loss": 4.6108,
+      "step": 12631
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001685387062738241,
+      "loss": 2.5227,
+      "step": 12632
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001686435727572297,
+      "loss": 3.9804,
+      "step": 12633
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001687482966842866,
+      "loss": 1.7101,
+      "step": 12634
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016885287783750862,
+      "loss": 0.573,
+      "step": 12635
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001689573159997051,
+      "loss": 0.3924,
+      "step": 12636
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016906161095398337,
+      "loss": 2.2085,
+      "step": 12637
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016916576248374713,
+      "loss": 1.7023,
+      "step": 12638
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016926977037269903,
+      "loss": 0.4563,
+      "step": 12639
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001693736344048389,
+      "loss": 2.4658,
+      "step": 12640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016947735436446642,
+      "loss": 0.5014,
+      "step": 12641
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001695809300361794,
+      "loss": 4.7301,
+      "step": 12642
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016968436120487642,
+      "loss": 3.2957,
+      "step": 12643
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000169787647655755,
+      "loss": 2.3881,
+      "step": 12644
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00016989078917431405,
+      "loss": 1.2605,
+      "step": 12645
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001699937855463535,
+      "loss": 2.5855,
+      "step": 12646
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017009663655797387,
+      "loss": 2.8996,
+      "step": 12647
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001701993419955784,
+      "loss": 2.5542,
+      "step": 12648
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017030190164587264,
+      "loss": 0.8807,
+      "step": 12649
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017040431529586402,
+      "loss": 2.346,
+      "step": 12650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001705065827328644,
+      "loss": 5.8331,
+      "step": 12651
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017060870374448813,
+      "loss": 0.7165,
+      "step": 12652
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017071067811865465,
+      "loss": 1.6445,
+      "step": 12653
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017081250564358728,
+      "loss": 1.6242,
+      "step": 12654
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017091418610781428,
+      "loss": 3.5982,
+      "step": 12655
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017101571930017,
+      "loss": 2.412,
+      "step": 12656
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001711171050097937,
+      "loss": 2.5277,
+      "step": 12657
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001712183430261319,
+      "loss": 1.9634,
+      "step": 12658
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001713194331389369,
+      "loss": 2.8482,
+      "step": 12659
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017142037513826906,
+      "loss": 0.547,
+      "step": 12660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017152116881449545,
+      "loss": 3.1019,
+      "step": 12661
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001716218139582921,
+      "loss": 3.9903,
+      "step": 12662
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017172231036064248,
+      "loss": 1.4579,
+      "step": 12663
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017182265781284003,
+      "loss": 3.4399,
+      "step": 12664
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001719228561064865,
+      "loss": 1.9605,
+      "step": 12665
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017202290503349444,
+      "loss": 0.7035,
+      "step": 12666
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001721228043860855,
+      "loss": 3.5082,
+      "step": 12667
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017222255395679304,
+      "loss": 0.591,
+      "step": 12668
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001723221535384605,
+      "loss": 2.1776,
+      "step": 12669
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017242160292424367,
+      "loss": 1.1647,
+      "step": 12670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001725209019076095,
+      "loss": 2.1832,
+      "step": 12671
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017262005028233812,
+      "loss": 2.787,
+      "step": 12672
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001727190478425214,
+      "loss": 3.6302,
+      "step": 12673
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017281789438256556,
+      "loss": 1.4526,
+      "step": 12674
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001729165896971894,
+      "loss": 1.7933,
+      "step": 12675
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017301513358142643,
+      "loss": 1.9009,
+      "step": 12676
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001731135258306248,
+      "loss": 1.2328,
+      "step": 12677
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017321176624044674,
+      "loss": 0.7596,
+      "step": 12678
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001733098546068708,
+      "loss": 0.2868,
+      "step": 12679
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017340779072619032,
+      "loss": 1.0444,
+      "step": 12680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017350557439501553,
+      "loss": 1.0002,
+      "step": 12681
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017360320541027333,
+      "loss": 1.7135,
+      "step": 12682
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017370068356920685,
+      "loss": 0.9341,
+      "step": 12683
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017379800866937765,
+      "loss": 2.5399,
+      "step": 12684
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017389518050866422,
+      "loss": 2.1577,
+      "step": 12685
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017399219888526428,
+      "loss": 3.999,
+      "step": 12686
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017408906359769354,
+      "loss": 3.7035,
+      "step": 12687
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001741857744447868,
+      "loss": 1.1641,
+      "step": 12688
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017428233122569915,
+      "loss": 2.4301,
+      "step": 12689
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017437873373990468,
+      "loss": 0.7815,
+      "step": 12690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017447498178719883,
+      "loss": 0.2943,
+      "step": 12691
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017457107516769688,
+      "loss": 1.7462,
+      "step": 12692
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017466701368183617,
+      "loss": 1.6592,
+      "step": 12693
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017476279713037486,
+      "loss": 1.8114,
+      "step": 12694
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017485842531439396,
+      "loss": 3.1039,
+      "step": 12695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017495389803529607,
+      "loss": 3.5027,
+      "step": 12696
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001750492150948076,
+      "loss": 0.8365,
+      "step": 12697
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017514437629497715,
+      "loss": 2.68,
+      "step": 12698
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017523938143817813,
+      "loss": 0.7125,
+      "step": 12699
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017533423032710702,
+      "loss": 2.0986,
+      "step": 12700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001754289227647857,
+      "loss": 2.9301,
+      "step": 12701
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017552345855456012,
+      "loss": 2.998,
+      "step": 12702
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017561783750010237,
+      "loss": 1.0101,
+      "step": 12703
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017571205940540945,
+      "loss": 1.1877,
+      "step": 12704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001758061240748053,
+      "loss": 1.0096,
+      "step": 12705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017590003131293965,
+      "loss": 2.929,
+      "step": 12706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017599378092478992,
+      "loss": 1.2023,
+      "step": 12707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017608737271566005,
+      "loss": 2.9619,
+      "step": 12708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017618080649118241,
+      "loss": 1.8644,
+      "step": 12709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017627408205731762,
+      "loss": 0.1723,
+      "step": 12710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017636719922035397,
+      "loss": 1.9669,
+      "step": 12711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001764601577869096,
+      "loss": 1.7774,
+      "step": 12712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001765529575639319,
+      "loss": 1.577,
+      "step": 12713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001766455983586974,
+      "loss": 3.8177,
+      "step": 12714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001767380799788137,
+      "loss": 0.7935,
+      "step": 12715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017683040223221807,
+      "loss": 2.9134,
+      "step": 12716
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001769225649271796,
+      "loss": 1.7361,
+      "step": 12717
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001770145678722981,
+      "loss": 2.2755,
+      "step": 12718
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017710641087650514,
+      "loss": 2.2616,
+      "step": 12719
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017719809374906514,
+      "loss": 4.6503,
+      "step": 12720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017728961629957422,
+      "loss": 1.0086,
+      "step": 12721
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017738097833796222,
+      "loss": 0.2626,
+      "step": 12722
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001774721796744915,
+      "loss": 2.1003,
+      "step": 12723
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001775632201197591,
+      "loss": 3.2007,
+      "step": 12724
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017765409948469524,
+      "loss": 0.3198,
+      "step": 12725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001777448175805656,
+      "loss": 3.3775,
+      "step": 12726
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001778353742189698,
+      "loss": 3.8431,
+      "step": 12727
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017792576921184377,
+      "loss": 1.3842,
+      "step": 12728
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017801600237145817,
+      "loss": 1.3049,
+      "step": 12729
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017810607351042068,
+      "loss": 0.8575,
+      "step": 12730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017819598244167456,
+      "loss": 0.805,
+      "step": 12731
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017828572897850083,
+      "loss": 2.5473,
+      "step": 12732
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017837531293451683,
+      "loss": 1.4141,
+      "step": 12733
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001784647341236785,
+      "loss": 3.5586,
+      "step": 12734
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001785539923602789,
+      "loss": 0.4294,
+      "step": 12735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017864308745895032,
+      "loss": 2.8713,
+      "step": 12736
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001787320192346631,
+      "loss": 3.9575,
+      "step": 12737
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017882078750272752,
+      "loss": 0.3581,
+      "step": 12738
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001789093920787926,
+      "loss": 2.702,
+      "step": 12739
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017899783277884796,
+      "loss": 0.8659,
+      "step": 12740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017908610941922359,
+      "loss": 2.4041,
+      "step": 12741
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017917422181658945,
+      "loss": 1.8082,
+      "step": 12742
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017926216978795729,
+      "loss": 0.6835,
+      "step": 12743
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001793499531506804,
+      "loss": 1.6416,
+      "step": 12744
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001794375717224532,
+      "loss": 3.7534,
+      "step": 12745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017952502532131317,
+      "loss": 3.2078,
+      "step": 12746
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017961231376563964,
+      "loss": 2.0614,
+      "step": 12747
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017969943687415568,
+      "loss": 2.2899,
+      "step": 12748
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017978639446592697,
+      "loss": 0.9497,
+      "step": 12749
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017987318636036374,
+      "loss": 0.9888,
+      "step": 12750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00017995981237721966,
+      "loss": 2.9467,
+      "step": 12751
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001800462723365929,
+      "loss": 0.279,
+      "step": 12752
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018013256605892722,
+      "loss": 2.8432,
+      "step": 12753
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001802186933650106,
+      "loss": 2.8168,
+      "step": 12754
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001803046540759776,
+      "loss": 2.3504,
+      "step": 12755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001803904480133079,
+      "loss": 2.8134,
+      "step": 12756
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001804760749988283,
+      "loss": 2.1358,
+      "step": 12757
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018056153485471158,
+      "loss": 1.7015,
+      "step": 12758
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001806468274034784,
+      "loss": 4.1917,
+      "step": 12759
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018073195246799602,
+      "loss": 2.8855,
+      "step": 12760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018081690987148034,
+      "loss": 1.9545,
+      "step": 12761
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001809016994374947,
+      "loss": 3.4974,
+      "step": 12762
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018098632098995177,
+      "loss": 0.7109,
+      "step": 12763
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018107077435311237,
+      "loss": 2.1117,
+      "step": 12764
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018115505935158732,
+      "loss": 0.5844,
+      "step": 12765
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018123917581033637,
+      "loss": 2.2501,
+      "step": 12766
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018132312355467013,
+      "loss": 3.1814,
+      "step": 12767
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018140690241024866,
+      "loss": 0.5144,
+      "step": 12768
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001814905122030837,
+      "loss": 1.6296,
+      "step": 12769
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001815739527595372,
+      "loss": 1.6589,
+      "step": 12770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018165722390632317,
+      "loss": 2.9726,
+      "step": 12771
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001817403254705075,
+      "loss": 0.6874,
+      "step": 12772
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018182325727950769,
+      "loss": 1.2598,
+      "step": 12773
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018190601916109444,
+      "loss": 0.6419,
+      "step": 12774
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001819886109433907,
+      "loss": 0.8029,
+      "step": 12775
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018207103245487317,
+      "loss": 0.9203,
+      "step": 12776
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018215328352437222,
+      "loss": 0.6664,
+      "step": 12777
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018223536398107157,
+      "loss": 5.5189,
+      "step": 12778
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001823172736545101,
+      "loss": 1.2131,
+      "step": 12779
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018239901237458045,
+      "loss": 0.4092,
+      "step": 12780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018248057997153128,
+      "loss": 2.9888,
+      "step": 12781
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018256197627596583,
+      "loss": 1.0057,
+      "step": 12782
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001826432011188432,
+      "loss": 1.609,
+      "step": 12783
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018272425433147918,
+      "loss": 1.5632,
+      "step": 12784
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018280513574554506,
+      "loss": 2.7785,
+      "step": 12785
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018288584519306992,
+      "loss": 1.0964,
+      "step": 12786
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000182966382506439,
+      "loss": 1.1667,
+      "step": 12787
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018304674751839587,
+      "loss": 2.0318,
+      "step": 12788
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018312694006204113,
+      "loss": 4.0659,
+      "step": 12789
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018320695997083448,
+      "loss": 2.3507,
+      "step": 12790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018328680707859315,
+      "loss": 2.3839,
+      "step": 12791
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018336648121949412,
+      "loss": 1.6311,
+      "step": 12792
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001834459822280728,
+      "loss": 3.6828,
+      "step": 12793
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001835253099392249,
+      "loss": 2.6952,
+      "step": 12794
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001836044641882053,
+      "loss": 4.6905,
+      "step": 12795
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018368344481062984,
+      "loss": 2.1714,
+      "step": 12796
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001837622516424742,
+      "loss": 1.1661,
+      "step": 12797
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018384088452007584,
+      "loss": 3.6553,
+      "step": 12798
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018391934328013248,
+      "loss": 2.9893,
+      "step": 12799
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018399762775970447,
+      "loss": 1.5517,
+      "step": 12800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001840757377962131,
+      "loss": 2.738,
+      "step": 12801
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018415367322744268,
+      "loss": 2.8646,
+      "step": 12802
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018423143389154008,
+      "loss": 3.1255,
+      "step": 12803
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018430901962701455,
+      "loss": 2.373,
+      "step": 12804
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018438643027273935,
+      "loss": 1.8295,
+      "step": 12805
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018446366566795066,
+      "loss": 3.392,
+      "step": 12806
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001845407256522491,
+      "loss": 1.6555,
+      "step": 12807
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018461761006559972,
+      "loss": 3.7956,
+      "step": 12808
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018469431874833147,
+      "loss": 1.8742,
+      "step": 12809
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018477085154113907,
+      "loss": 2.3555,
+      "step": 12810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018484720828508176,
+      "loss": 2.5717,
+      "step": 12811
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018492338882158515,
+      "loss": 2.6785,
+      "step": 12812
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018499939299244006,
+      "loss": 1.876,
+      "step": 12813
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001850752206398043,
+      "loss": 0.6301,
+      "step": 12814
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018515087160620168,
+      "loss": 3.8979,
+      "step": 12815
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018522634573452305,
+      "loss": 2.1307,
+      "step": 12816
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018530164286802695,
+      "loss": 2.0316,
+      "step": 12817
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018537676285033881,
+      "loss": 0.69,
+      "step": 12818
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018545170552545267,
+      "loss": 0.9999,
+      "step": 12819
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018552647073773014,
+      "loss": 0.7573,
+      "step": 12820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018560105833190197,
+      "loss": 2.8554,
+      "step": 12821
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001856754681530672,
+      "loss": 4.0831,
+      "step": 12822
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018574970004669468,
+      "loss": 1.6058,
+      "step": 12823
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018582375385862212,
+      "loss": 0.4777,
+      "step": 12824
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001858976294350577,
+      "loss": 2.5913,
+      "step": 12825
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018597132662257913,
+      "loss": 3.0469,
+      "step": 12826
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018604484526813528,
+      "loss": 3.6046,
+      "step": 12827
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018611818521904505,
+      "loss": 1.6989,
+      "step": 12828
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001861913463229992,
+      "loss": 0.7069,
+      "step": 12829
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018626432842805917,
+      "loss": 1.6481,
+      "step": 12830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001863371313826589,
+      "loss": 2.444,
+      "step": 12831
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001864097550356036,
+      "loss": 2.1744,
+      "step": 12832
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018648219923607165,
+      "loss": 2.1403,
+      "step": 12833
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018655446383361327,
+      "loss": 2.6515,
+      "step": 12834
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001866265486781524,
+      "loss": 1.0664,
+      "step": 12835
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018669845361998606,
+      "loss": 6.2512,
+      "step": 12836
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001867701785097845,
+      "loss": 1.4115,
+      "step": 12837
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018684172319859244,
+      "loss": 1.7743,
+      "step": 12838
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018691308753782866,
+      "loss": 2.746,
+      "step": 12839
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018698427137928621,
+      "loss": 1.3881,
+      "step": 12840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018705527457513345,
+      "loss": 0.6753,
+      "step": 12841
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018712609697791335,
+      "loss": 2.3892,
+      "step": 12842
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018719673844054503,
+      "loss": 2.2308,
+      "step": 12843
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018726719881632258,
+      "loss": 0.9935,
+      "step": 12844
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018733747795891703,
+      "loss": 2.2597,
+      "step": 12845
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018740757572237508,
+      "loss": 0.9936,
+      "step": 12846
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018747749196112025,
+      "loss": 2.6289,
+      "step": 12847
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018754722652995353,
+      "loss": 1.7625,
+      "step": 12848
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001876167792840524,
+      "loss": 2.3268,
+      "step": 12849
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018768615007897277,
+      "loss": 0.7451,
+      "step": 12850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018775533877064766,
+      "loss": 1.154,
+      "step": 12851
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018782434521538903,
+      "loss": 0.9783,
+      "step": 12852
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001878931692698865,
+      "loss": 2.8011,
+      "step": 12853
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018796181079120934,
+      "loss": 3.8289,
+      "step": 12854
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018803026963680507,
+      "loss": 0.195,
+      "step": 12855
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001880985456645013,
+      "loss": 1.0503,
+      "step": 12856
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018816663873250471,
+      "loss": 0.7802,
+      "step": 12857
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001882345486994025,
+      "loss": 3.5575,
+      "step": 12858
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001883022754241615,
+      "loss": 2.7146,
+      "step": 12859
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018836981876612984,
+      "loss": 1.2043,
+      "step": 12860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001884371785850357,
+      "loss": 3.651,
+      "step": 12861
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018850435474098914,
+      "loss": 4.3245,
+      "step": 12862
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001885713470944809,
+      "loss": 2.9245,
+      "step": 12863
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018863815550638422,
+      "loss": 0.2956,
+      "step": 12864
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018870477983795359,
+      "loss": 2.3266,
+      "step": 12865
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018877121995082628,
+      "loss": 1.2856,
+      "step": 12866
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018883747570702224,
+      "loss": 4.2477,
+      "step": 12867
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018890354696894367,
+      "loss": 3.229,
+      "step": 12868
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018896943359937644,
+      "loss": 4.0406,
+      "step": 12869
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018903513546148992,
+      "loss": 1.4777,
+      "step": 12870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018910065241883666,
+      "loss": 3.8859,
+      "step": 12871
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018916598433535385,
+      "loss": 1.0486,
+      "step": 12872
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018923113107536233,
+      "loss": 2.494,
+      "step": 12873
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018929609250356819,
+      "loss": 3.1657,
+      "step": 12874
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018936086848506172,
+      "loss": 1.8851,
+      "step": 12875
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018942545888531891,
+      "loss": 0.6142,
+      "step": 12876
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001894898635702006,
+      "loss": 3.0105,
+      "step": 12877
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001895540824059539,
+      "loss": 3.3065,
+      "step": 12878
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001896181152592114,
+      "loss": 1.8265,
+      "step": 12879
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018968196199699202,
+      "loss": 0.2492,
+      "step": 12880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001897456224867016,
+      "loss": 2.1068,
+      "step": 12881
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018980909659613212,
+      "loss": 2.7223,
+      "step": 12882
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001898723841934633,
+      "loss": 2.3511,
+      "step": 12883
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018993548514726157,
+      "loss": 1.4055,
+      "step": 12884
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001899983993264816,
+      "loss": 2.0626,
+      "step": 12885
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019006112660046532,
+      "loss": 1.8344,
+      "step": 12886
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019012366683894338,
+      "loss": 3.3594,
+      "step": 12887
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019018601991203435,
+      "loss": 3.6961,
+      "step": 12888
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019024818569024607,
+      "loss": 3.878,
+      "step": 12889
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019031016404447469,
+      "loss": 0.6084,
+      "step": 12890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019037195484600624,
+      "loss": 0.4309,
+      "step": 12891
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019043355796651563,
+      "loss": 2.5067,
+      "step": 12892
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019049497327806817,
+      "loss": 1.7147,
+      "step": 12893
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019055620065311861,
+      "loss": 1.7745,
+      "step": 12894
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019061723996451256,
+      "loss": 3.692,
+      "step": 12895
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019067809108548557,
+      "loss": 2.3308,
+      "step": 12896
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001907387538896646,
+      "loss": 0.7817,
+      "step": 12897
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001907992282510675,
+      "loss": 4.5078,
+      "step": 12898
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001908595140441032,
+      "loss": 2.324,
+      "step": 12899
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019091961114357274,
+      "loss": 0.5988,
+      "step": 12900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019097951942466845,
+      "loss": 2.2945,
+      "step": 12901
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019103923876297532,
+      "loss": 3.3702,
+      "step": 12902
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019109876903447058,
+      "loss": 2.4292,
+      "step": 12903
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019115811011552377,
+      "loss": 0.7363,
+      "step": 12904
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001912172618828979,
+      "loss": 2.2614,
+      "step": 12905
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019127622421374852,
+      "loss": 2.102,
+      "step": 12906
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019133499698562525,
+      "loss": 4.2889,
+      "step": 12907
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019139358007647073,
+      "loss": 3.171,
+      "step": 12908
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019145197336462212,
+      "loss": 1.0285,
+      "step": 12909
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019151017672881034,
+      "loss": 3.1619,
+      "step": 12910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001915681900481607,
+      "loss": 1.7413,
+      "step": 12911
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019162601320219368,
+      "loss": 0.64,
+      "step": 12912
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019168364607082406,
+      "loss": 2.9846,
+      "step": 12913
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019174108853436237,
+      "loss": 1.7263,
+      "step": 12914
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019179834047351402,
+      "loss": 3.1056,
+      "step": 12915
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019185540176938068,
+      "loss": 1.6244,
+      "step": 12916
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001919122723034593,
+      "loss": 0.2445,
+      "step": 12917
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019196895195764367,
+      "loss": 3.6976,
+      "step": 12918
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019202544061422332,
+      "loss": 1.4605,
+      "step": 12919
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019208173815588508,
+      "loss": 0.1907,
+      "step": 12920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019213784446571209,
+      "loss": 2.431,
+      "step": 12921
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019219375942718513,
+      "loss": 1.2306,
+      "step": 12922
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019224948292418192,
+      "loss": 2.1044,
+      "step": 12923
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001923050148409783,
+      "loss": 1.7056,
+      "step": 12924
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019236035506224743,
+      "loss": 2.8562,
+      "step": 12925
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019241550347306115,
+      "loss": 1.2581,
+      "step": 12926
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019247045995888904,
+      "loss": 2.375,
+      "step": 12927
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019252522440559985,
+      "loss": 2.4039,
+      "step": 12928
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019257979669946057,
+      "loss": 0.6486,
+      "step": 12929
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019263417672713774,
+      "loss": 2.6673,
+      "step": 12930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019268836437569712,
+      "loss": 1.5221,
+      "step": 12931
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001927423595326036,
+      "loss": 3.4355,
+      "step": 12932
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001927961620857222,
+      "loss": 0.941,
+      "step": 12933
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000192849771923318,
+      "loss": 4.1049,
+      "step": 12934
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019290318893405587,
+      "loss": 4.0164,
+      "step": 12935
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019295641300700166,
+      "loss": 2.8219,
+      "step": 12936
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019300944403162138,
+      "loss": 0.4533,
+      "step": 12937
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001930622818977825,
+      "loss": 2.0032,
+      "step": 12938
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001931149264957531,
+      "loss": 1.7492,
+      "step": 12939
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019316737771620313,
+      "loss": 3.1009,
+      "step": 12940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019321963545020367,
+      "loss": 2.2699,
+      "step": 12941
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001932716995892281,
+      "loss": 3.9196,
+      "step": 12942
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019332357002515148,
+      "loss": 2.163,
+      "step": 12943
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001933752466502511,
+      "loss": 2.1794,
+      "step": 12944
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019342672935720727,
+      "loss": 0.7341,
+      "step": 12945
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019347801803910235,
+      "loss": 3.7798,
+      "step": 12946
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019352911258942222,
+      "loss": 3.3322,
+      "step": 12947
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001935800129020554,
+      "loss": 3.2333,
+      "step": 12948
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001936307188712944,
+      "loss": 1.2977,
+      "step": 12949
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019368123039183464,
+      "loss": 5.3658,
+      "step": 12950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019373154735877608,
+      "loss": 4.706,
+      "step": 12951
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019378166966762205,
+      "loss": 1.3845,
+      "step": 12952
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001938315972142807,
+      "loss": 1.4993,
+      "step": 12953
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019388132989506422,
+      "loss": 2.7754,
+      "step": 12954
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019393086760668987,
+      "loss": 2.9327,
+      "step": 12955
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001939802102462794,
+      "loss": 1.9531,
+      "step": 12956
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019402935771136015,
+      "loss": 0.4363,
+      "step": 12957
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001940783098998643,
+      "loss": 0.4933,
+      "step": 12958
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019412706671013003,
+      "loss": 2.364,
+      "step": 12959
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019417562804090082,
+      "loss": 1.5358,
+      "step": 12960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019422399379132653,
+      "loss": 0.1899,
+      "step": 12961
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019427216386096313,
+      "loss": 1.3102,
+      "step": 12962
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019432013814977252,
+      "loss": 1.8372,
+      "step": 12963
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019436791655812375,
+      "loss": 3.9713,
+      "step": 12964
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019441549898679256,
+      "loss": 0.634,
+      "step": 12965
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001944628853369613,
+      "loss": 2.4259,
+      "step": 12966
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001945100755102201,
+      "loss": 2.8065,
+      "step": 12967
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019455706940856593,
+      "loss": 3.4799,
+      "step": 12968
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019460386693440388,
+      "loss": 1.5391,
+      "step": 12969
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019465046799054645,
+      "loss": 2.5947,
+      "step": 12970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019469687248021455,
+      "loss": 1.6193,
+      "step": 12971
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019474308030703692,
+      "loss": 0.8405,
+      "step": 12972
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001947890913750511,
+      "loss": 1.0451,
+      "step": 12973
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019483490558870292,
+      "loss": 1.4337,
+      "step": 12974
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019488052285284712,
+      "loss": 3.2485,
+      "step": 12975
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001949259430727477,
+      "loss": 2.9442,
+      "step": 12976
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019497116615407748,
+      "loss": 0.4317,
+      "step": 12977
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019501619200291913,
+      "loss": 0.8043,
+      "step": 12978
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001950610205257644,
+      "loss": 2.9484,
+      "step": 12979
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001951056516295154,
+      "loss": 6.0351,
+      "step": 12980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019515008522148373,
+      "loss": 1.5491,
+      "step": 12981
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019519432120939162,
+      "loss": 0.5223,
+      "step": 12982
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019523835950137125,
+      "loss": 1.0707,
+      "step": 12983
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001952822000059658,
+      "loss": 1.3649,
+      "step": 12984
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019532584263212878,
+      "loss": 2.2551,
+      "step": 12985
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000195369287289225,
+      "loss": 3.108,
+      "step": 12986
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001954125338870301,
+      "loss": 3.1763,
+      "step": 12987
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001954555823357314,
+      "loss": 1.86,
+      "step": 12988
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019549843254592736,
+      "loss": 1.5021,
+      "step": 12989
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019554108442862843,
+      "loss": 2.6588,
+      "step": 12990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019558353789525662,
+      "loss": 0.9095,
+      "step": 12991
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019562579285764632,
+      "loss": 2.3752,
+      "step": 12992
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019566784922804418,
+      "loss": 1.2146,
+      "step": 12993
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019570970691910885,
+      "loss": 2.2779,
+      "step": 12994
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019575136584391203,
+      "loss": 2.6153,
+      "step": 12995
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019579282591593815,
+      "loss": 1.7295,
+      "step": 12996
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001958340870490843,
+      "loss": 2.6709,
+      "step": 12997
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001958751491576612,
+      "loss": 2.661,
+      "step": 12998
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019591601215639232,
+      "loss": 2.007,
+      "step": 12999
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019595667596041525,
+      "loss": 1.9187,
+      "step": 13000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001959971404852807,
+      "loss": 3.0924,
+      "step": 13001
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001960374056469537,
+      "loss": 1.5186,
+      "step": 13002
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001960774713618129,
+      "loss": 1.4315,
+      "step": 13003
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019611733754665157,
+      "loss": 3.2435,
+      "step": 13004
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019615700411867692,
+      "loss": 3.3074,
+      "step": 13005
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019619647099551114,
+      "loss": 0.6735,
+      "step": 13006
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019623573809519082,
+      "loss": 2.5271,
+      "step": 13007
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019627480533616746,
+      "loss": 0.4131,
+      "step": 13008
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001963136726373079,
+      "loss": 1.4732,
+      "step": 13009
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019635233991789383,
+      "loss": 0.9764,
+      "step": 13010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001963908070976227,
+      "loss": 3.6575,
+      "step": 13011
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019642907409660719,
+      "loss": 4.6089,
+      "step": 13012
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019646714083537597,
+      "loss": 2.555,
+      "step": 13013
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019650500723487335,
+      "loss": 2.4723,
+      "step": 13014
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019654267321646003,
+      "loss": 3.9636,
+      "step": 13015
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019658013870191255,
+      "loss": 3.6882,
+      "step": 13016
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019661740361342418,
+      "loss": 1.496,
+      "step": 13017
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001966544678736044,
+      "loss": 2.7814,
+      "step": 13018
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001966913314054798,
+      "loss": 2.2641,
+      "step": 13019
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001967279941324934,
+      "loss": 2.1171,
+      "step": 13020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001967644559785056,
+      "loss": 3.2229,
+      "step": 13021
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019680071686779366,
+      "loss": 3.2867,
+      "step": 13022
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019683677672505248,
+      "loss": 3.4437,
+      "step": 13023
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019687263547539437,
+      "loss": 2.221,
+      "step": 13024
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001969082930443491,
+      "loss": 1.4068,
+      "step": 13025
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019694374935786457,
+      "loss": 1.6917,
+      "step": 13026
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019697900434230629,
+      "loss": 1.2312,
+      "step": 13027
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019701405792445808,
+      "loss": 1.6924,
+      "step": 13028
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019704891003152216,
+      "loss": 2.7337,
+      "step": 13029
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001970835605911188,
+      "loss": 3.6866,
+      "step": 13030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019711800953128728,
+      "loss": 1.5999,
+      "step": 13031
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019715225678048506,
+      "loss": 3.2878,
+      "step": 13032
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019718630226758904,
+      "loss": 1.4409,
+      "step": 13033
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019722014592189464,
+      "loss": 3.6141,
+      "step": 13034
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019725378767311683,
+      "loss": 2.7096,
+      "step": 13035
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019728722745138955,
+      "loss": 1.1291,
+      "step": 13036
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001973204651872665,
+      "loss": 2.6494,
+      "step": 13037
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001973535008117207,
+      "loss": 1.4178,
+      "step": 13038
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000197386334256145,
+      "loss": 2.0448,
+      "step": 13039
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001974189654523523,
+      "loss": 1.6792,
+      "step": 13040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019745139433257524,
+      "loss": 0.8689,
+      "step": 13041
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001974836208294669,
+      "loss": 0.6915,
+      "step": 13042
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019751564487610038,
+      "loss": 5.5853,
+      "step": 13043
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019754746640596947,
+      "loss": 1.2895,
+      "step": 13044
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019757908535298832,
+      "loss": 3.4889,
+      "step": 13045
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019761050165149208,
+      "loss": 2.9361,
+      "step": 13046
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019764171523623637,
+      "loss": 2.8089,
+      "step": 13047
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019767272604239824,
+      "loss": 2.2327,
+      "step": 13048
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019770353400557544,
+      "loss": 3.832,
+      "step": 13049
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019773413906178731,
+      "loss": 1.6462,
+      "step": 13050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001977645411474743,
+      "loss": 0.2609,
+      "step": 13051
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019779474019949867,
+      "loss": 0.3134,
+      "step": 13052
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019782473615514404,
+      "loss": 1.2448,
+      "step": 13053
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001978545289521161,
+      "loss": 0.455,
+      "step": 13054
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019788411852854213,
+      "loss": 2.9435,
+      "step": 13055
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019791350482297176,
+      "loss": 1.3178,
+      "step": 13056
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019794268777437667,
+      "loss": 5.396,
+      "step": 13057
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019797166732215074,
+      "loss": 3.5196,
+      "step": 13058
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019800044340611037,
+      "loss": 1.5857,
+      "step": 13059
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001980290159664946,
+      "loss": 1.2152,
+      "step": 13060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019805738494396485,
+      "loss": 1.0894,
+      "step": 13061
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019808555027960565,
+      "loss": 3.5943,
+      "step": 13062
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019811351191492422,
+      "loss": 3.2127,
+      "step": 13063
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000198141269791851,
+      "loss": 4.5644,
+      "step": 13064
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019816882385273937,
+      "loss": 1.9818,
+      "step": 13065
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019819617404036625,
+      "loss": 4.0165,
+      "step": 13066
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019822332029793168,
+      "loss": 2.6001,
+      "step": 13067
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019825026256905945,
+      "loss": 2.584,
+      "step": 13068
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001982770007977968,
+      "loss": 1.6115,
+      "step": 13069
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019830353492861487,
+      "loss": 2.4815,
+      "step": 13070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019832986490640856,
+      "loss": 3.2926,
+      "step": 13071
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019835599067649675,
+      "loss": 4.4521,
+      "step": 13072
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001983819121846225,
+      "loss": 2.3051,
+      "step": 13073
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019840762937695295,
+      "loss": 1.4261,
+      "step": 13074
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001984331422000797,
+      "loss": 3.3143,
+      "step": 13075
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019845845060101856,
+      "loss": 3.5415,
+      "step": 13076
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019848355452721018,
+      "loss": 2.0058,
+      "step": 13077
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001985084539265195,
+      "loss": 4.2315,
+      "step": 13078
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019853314874723657,
+      "loss": 1.9498,
+      "step": 13079
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000198557638938076,
+      "loss": 1.6214,
+      "step": 13080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019858192444817762,
+      "loss": 2.2298,
+      "step": 13081
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001986060052271061,
+      "loss": 0.9549,
+      "step": 13082
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019862988122485147,
+      "loss": 1.928,
+      "step": 13083
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001986535523918289,
+      "loss": 2.6664,
+      "step": 13084
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001986770186788791,
+      "loss": 1.8971,
+      "step": 13085
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001987002800372681,
+      "loss": 2.2502,
+      "step": 13086
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019872333641868765,
+      "loss": 3.165,
+      "step": 13087
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019874618777525517,
+      "loss": 0.6648,
+      "step": 13088
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019876883405951375,
+      "loss": 1.1439,
+      "step": 13089
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001987912752244325,
+      "loss": 2.3852,
+      "step": 13090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001988135112234065,
+      "loss": 0.589,
+      "step": 13091
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019883554201025682,
+      "loss": 1.8413,
+      "step": 13092
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001988573675392308,
+      "loss": 3.0487,
+      "step": 13093
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019887898776500197,
+      "loss": 3.5659,
+      "step": 13094
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019890040264267035,
+      "loss": 1.7209,
+      "step": 13095
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019892161212776226,
+      "loss": 2.4236,
+      "step": 13096
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019894261617623072,
+      "loss": 2.7199,
+      "step": 13097
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019896341474445522,
+      "loss": 3.3565,
+      "step": 13098
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001989840077892422,
+      "loss": 1.3835,
+      "step": 13099
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019900439526782476,
+      "loss": 2.4729,
+      "step": 13100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019902457713786298,
+      "loss": 3.5365,
+      "step": 13101
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019904455335744396,
+      "loss": 1.3478,
+      "step": 13102
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019906432388508175,
+      "loss": 1.2463,
+      "step": 13103
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019908388867971775,
+      "loss": 2.242,
+      "step": 13104
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019910324770072048,
+      "loss": 1.8719,
+      "step": 13105
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019912240090788597,
+      "loss": 2.1068,
+      "step": 13106
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019914134826143744,
+      "loss": 2.4956,
+      "step": 13107
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001991600897220259,
+      "loss": 0.8797,
+      "step": 13108
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019917862525072966,
+      "loss": 2.45,
+      "step": 13109
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019919695480905496,
+      "loss": 2.1121,
+      "step": 13110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019921507835893553,
+      "loss": 4.6356,
+      "step": 13111
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019923299586273324,
+      "loss": 1.2612,
+      "step": 13112
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019925070728323758,
+      "loss": 1.4491,
+      "step": 13113
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019926821258366622,
+      "loss": 3.6816,
+      "step": 13114
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019928551172766477,
+      "loss": 2.0727,
+      "step": 13115
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019930260467930708,
+      "loss": 0.2589,
+      "step": 13116
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001993194914030951,
+      "loss": 1.3148,
+      "step": 13117
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019933617186395917,
+      "loss": 3.4905,
+      "step": 13118
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019935264602725797,
+      "loss": 3.3926,
+      "step": 13119
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001993689138587785,
+      "loss": 1.1416,
+      "step": 13120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019938497532473646,
+      "loss": 1.2347,
+      "step": 13121
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001994008303917759,
+      "loss": 4.5486,
+      "step": 13122
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001994164790269697,
+      "loss": 1.1099,
+      "step": 13123
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019943192119781938,
+      "loss": 0.9569,
+      "step": 13124
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019944715687225518,
+      "loss": 1.4102,
+      "step": 13125
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019946218601863625,
+      "loss": 2.6182,
+      "step": 13126
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019947700860575062,
+      "loss": 4.2537,
+      "step": 13127
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001994916246028154,
+      "loss": 2.9379,
+      "step": 13128
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019950603397947647,
+      "loss": 2.8095,
+      "step": 13129
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019952023670580916,
+      "loss": 2.6947,
+      "step": 13130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019953423275231765,
+      "loss": 2.1941,
+      "step": 13131
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019954802208993564,
+      "loss": 1.371,
+      "step": 13132
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019956160469002582,
+      "loss": 4.7582,
+      "step": 13133
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019957498052438044,
+      "loss": 3.7088,
+      "step": 13134
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019958814956522108,
+      "loss": 2.0966,
+      "step": 13135
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001996011117851987,
+      "loss": 2.2809,
+      "step": 13136
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019961386715739397,
+      "loss": 2.5594,
+      "step": 13137
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019962641565531692,
+      "loss": 3.2947,
+      "step": 13138
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019963875725290738,
+      "loss": 0.4421,
+      "step": 13139
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019965089192453475,
+      "loss": 2.5119,
+      "step": 13140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019966281964499825,
+      "loss": 1.4998,
+      "step": 13141
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019967454038952677,
+      "loss": 3.4826,
+      "step": 13142
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019968605413377925,
+      "loss": 1.9848,
+      "step": 13143
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001996973608538442,
+      "loss": 4.4843,
+      "step": 13144
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019970846052624043,
+      "loss": 2.0347,
+      "step": 13145
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019971935312791645,
+      "loss": 0.805,
+      "step": 13146
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000199730038636251,
+      "loss": 1.833,
+      "step": 13147
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019974051702905277,
+      "loss": 1.2225,
+      "step": 13148
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019975078828456062,
+      "loss": 0.1965,
+      "step": 13149
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001997608523814437,
+      "loss": 3.2552,
+      "step": 13150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019977070929880114,
+      "loss": 3.3725,
+      "step": 13151
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001997803590161626,
+      "loss": 3.6574,
+      "step": 13152
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019978980151348782,
+      "loss": 1.0389,
+      "step": 13153
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000199799036771167,
+      "loss": 1.7034,
+      "step": 13154
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019980806477002078,
+      "loss": 0.7381,
+      "step": 13155
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019981688549130006,
+      "loss": 2.4421,
+      "step": 13156
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019982549891668639,
+      "loss": 3.0034,
+      "step": 13157
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019983390502829166,
+      "loss": 0.4795,
+      "step": 13158
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001998421038086584,
+      "loss": 2.722,
+      "step": 13159
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019985009524075978,
+      "loss": 0.3635,
+      "step": 13160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019985787930799939,
+      "loss": 1.6804,
+      "step": 13161
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001998654559942116,
+      "loss": 1.9454,
+      "step": 13162
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019987282528366148,
+      "loss": 1.9517,
+      "step": 13163
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001998799871610447,
+      "loss": 2.5873,
+      "step": 13164
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019988694161148787,
+      "loss": 1.1588,
+      "step": 13165
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019989368862054812,
+      "loss": 1.4756,
+      "step": 13166
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019990022817421358,
+      "loss": 2.0301,
+      "step": 13167
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019990656025890315,
+      "loss": 2.3047,
+      "step": 13168
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019991268486146662,
+      "loss": 1.3988,
+      "step": 13169
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999186019691846,
+      "loss": 3.3074,
+      "step": 13170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999243115697687,
+      "loss": 2.426,
+      "step": 13171
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019992981365136146,
+      "loss": 1.6129,
+      "step": 13172
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999351082025363,
+      "loss": 1.4334,
+      "step": 13173
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019994019521229776,
+      "loss": 2.6884,
+      "step": 13174
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019994507467008133,
+      "loss": 1.5437,
+      "step": 13175
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019994974656575352,
+      "loss": 1.0595,
+      "step": 13176
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999542108896119,
+      "loss": 1.3234,
+      "step": 13177
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019995846763238512,
+      "loss": 3.7702,
+      "step": 13178
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000199962516785233,
+      "loss": 1.6324,
+      "step": 13179
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019996635833974637,
+      "loss": 3.9235,
+      "step": 13180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019996999228794727,
+      "loss": 2.5706,
+      "step": 13181
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019997341862228882,
+      "loss": 2.6562,
+      "step": 13182
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019997663733565537,
+      "loss": 1.2715,
+      "step": 13183
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999796484213624,
+      "loss": 1.616,
+      "step": 13184
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999824518731566,
+      "loss": 2.2046,
+      "step": 13185
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019998504768521587,
+      "loss": 1.5733,
+      "step": 13186
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019998743585214932,
+      "loss": 3.8923,
+      "step": 13187
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019998961636899734,
+      "loss": 1.9508,
+      "step": 13188
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999158923123144,
+      "loss": 5.5324,
+      "step": 13189
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999335443475453,
+      "loss": 6.0842,
+      "step": 13190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999491197590066,
+      "loss": 0.8432,
+      "step": 13191
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999626185143518,
+      "loss": 2.2055,
+      "step": 13192
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999740405855478,
+      "loss": 1.8329,
+      "step": 13193
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999983385948873,
+      "loss": 2.2072,
+      "step": 13194
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999906545849197,
+      "loss": 5.0798,
+      "step": 13195
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999958464785923,
+      "loss": 2.7579,
+      "step": 13196
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999998961619109,
+      "loss": 1.6293,
+      "step": 13197
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 2.6738,
+      "step": 13198
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999998961619109,
+      "loss": 3.7214,
+      "step": 13199
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999958464785923,
+      "loss": 0.9894,
+      "step": 13200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999906545849197,
+      "loss": 0.2878,
+      "step": 13201
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999983385948873,
+      "loss": 1.7849,
+      "step": 13202
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999740405855478,
+      "loss": 2.5396,
+      "step": 13203
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999626185143518,
+      "loss": 0.9831,
+      "step": 13204
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999491197590066,
+      "loss": 1.4032,
+      "step": 13205
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999335443475453,
+      "loss": 4.9279,
+      "step": 13206
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019999158923123144,
+      "loss": 3.6996,
+      "step": 13207
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019998961636899734,
+      "loss": 0.6126,
+      "step": 13208
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019998743585214932,
+      "loss": 2.4014,
+      "step": 13209
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019998504768521587,
+      "loss": 0.7559,
+      "step": 13210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999824518731566,
+      "loss": 2.3534,
+      "step": 13211
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999796484213624,
+      "loss": 1.139,
+      "step": 13212
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019997663733565537,
+      "loss": 2.1231,
+      "step": 13213
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019997341862228884,
+      "loss": 3.375,
+      "step": 13214
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999699922879473,
+      "loss": 1.1295,
+      "step": 13215
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999663583397464,
+      "loss": 0.3319,
+      "step": 13216
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019996251678523302,
+      "loss": 2.4626,
+      "step": 13217
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019995846763238512,
+      "loss": 1.9819,
+      "step": 13218
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999542108896119,
+      "loss": 0.5999,
+      "step": 13219
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019994974656575352,
+      "loss": 2.2851,
+      "step": 13220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019994507467008133,
+      "loss": 1.7159,
+      "step": 13221
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019994019521229776,
+      "loss": 3.0838,
+      "step": 13222
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999351082025363,
+      "loss": 1.5226,
+      "step": 13223
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019992981365136146,
+      "loss": 2.5548,
+      "step": 13224
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001999243115697687,
+      "loss": 1.6815,
+      "step": 13225
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019991860196918462,
+      "loss": 3.228,
+      "step": 13226
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019991268486146662,
+      "loss": 1.3114,
+      "step": 13227
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019990656025890315,
+      "loss": 1.5483,
+      "step": 13228
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019990022817421358,
+      "loss": 1.8301,
+      "step": 13229
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019989368862054812,
+      "loss": 0.8261,
+      "step": 13230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019988694161148787,
+      "loss": 1.2874,
+      "step": 13231
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019987998716104472,
+      "loss": 2.496,
+      "step": 13232
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019987282528366148,
+      "loss": 1.7414,
+      "step": 13233
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001998654559942116,
+      "loss": 1.854,
+      "step": 13234
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019985787930799939,
+      "loss": 0.4596,
+      "step": 13235
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019985009524075978,
+      "loss": 0.3127,
+      "step": 13236
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001998421038086584,
+      "loss": 3.0623,
+      "step": 13237
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019983390502829166,
+      "loss": 0.6374,
+      "step": 13238
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019982549891668639,
+      "loss": 2.0321,
+      "step": 13239
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019981688549130006,
+      "loss": 1.8624,
+      "step": 13240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019980806477002078,
+      "loss": 1.5444,
+      "step": 13241
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019979903677116702,
+      "loss": 1.8198,
+      "step": 13242
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019978980151348782,
+      "loss": 1.1054,
+      "step": 13243
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001997803590161626,
+      "loss": 3.1273,
+      "step": 13244
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019977070929880116,
+      "loss": 2.1267,
+      "step": 13245
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001997608523814437,
+      "loss": 1.0163,
+      "step": 13246
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019975078828456065,
+      "loss": 3.0533,
+      "step": 13247
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019974051702905277,
+      "loss": 2.3747,
+      "step": 13248
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000199730038636251,
+      "loss": 2.2645,
+      "step": 13249
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019971935312791648,
+      "loss": 1.0106,
+      "step": 13250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019970846052624043,
+      "loss": 3.1147,
+      "step": 13251
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019969736085384424,
+      "loss": 1.9331,
+      "step": 13252
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019968605413377925,
+      "loss": 0.5785,
+      "step": 13253
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001996745403895268,
+      "loss": 3.1384,
+      "step": 13254
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019966281964499825,
+      "loss": 4.505,
+      "step": 13255
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019965089192453477,
+      "loss": 0.7285,
+      "step": 13256
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019963875725290738,
+      "loss": 0.2611,
+      "step": 13257
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019962641565531692,
+      "loss": 0.2703,
+      "step": 13258
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019961386715739397,
+      "loss": 1.3932,
+      "step": 13259
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001996011117851987,
+      "loss": 0.7566,
+      "step": 13260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019958814956522108,
+      "loss": 1.5676,
+      "step": 13261
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019957498052438046,
+      "loss": 3.213,
+      "step": 13262
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019956160469002582,
+      "loss": 3.5714,
+      "step": 13263
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019954802208993564,
+      "loss": 1.8139,
+      "step": 13264
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019953423275231768,
+      "loss": 3.336,
+      "step": 13265
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019952023670580918,
+      "loss": 2.2818,
+      "step": 13266
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001995060339794765,
+      "loss": 2.2163,
+      "step": 13267
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001994916246028154,
+      "loss": 4.273,
+      "step": 13268
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019947700860575065,
+      "loss": 0.7536,
+      "step": 13269
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019946218601863625,
+      "loss": 1.7262,
+      "step": 13270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019944715687225518,
+      "loss": 3.3198,
+      "step": 13271
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019943192119781938,
+      "loss": 3.3203,
+      "step": 13272
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001994164790269697,
+      "loss": 2.8079,
+      "step": 13273
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019940083039177593,
+      "loss": 3.5129,
+      "step": 13274
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019938497532473646,
+      "loss": 1.5135,
+      "step": 13275
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001993689138587785,
+      "loss": 0.9264,
+      "step": 13276
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019935264602725797,
+      "loss": 1.208,
+      "step": 13277
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019933617186395917,
+      "loss": 1.8128,
+      "step": 13278
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001993194914030951,
+      "loss": 0.3386,
+      "step": 13279
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001993026046793071,
+      "loss": 1.3122,
+      "step": 13280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019928551172766477,
+      "loss": 1.9901,
+      "step": 13281
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019926821258366622,
+      "loss": 3.3313,
+      "step": 13282
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001992507072832376,
+      "loss": 2.716,
+      "step": 13283
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019923299586273324,
+      "loss": 2.1936,
+      "step": 13284
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019921507835893556,
+      "loss": 1.3962,
+      "step": 13285
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019919695480905496,
+      "loss": 2.5711,
+      "step": 13286
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001991786252507297,
+      "loss": 0.5196,
+      "step": 13287
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001991600897220259,
+      "loss": 2.0392,
+      "step": 13288
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001991413482614375,
+      "loss": 1.9198,
+      "step": 13289
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000199122400907886,
+      "loss": 1.7835,
+      "step": 13290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001991032477007205,
+      "loss": 2.803,
+      "step": 13291
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019908388867971778,
+      "loss": 5.1068,
+      "step": 13292
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019906432388508175,
+      "loss": 3.3311,
+      "step": 13293
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019904455335744396,
+      "loss": 0.8297,
+      "step": 13294
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000199024577137863,
+      "loss": 2.3101,
+      "step": 13295
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019900439526782479,
+      "loss": 1.9875,
+      "step": 13296
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019898400778924224,
+      "loss": 1.9375,
+      "step": 13297
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019896341474445522,
+      "loss": 2.1118,
+      "step": 13298
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019894261617623072,
+      "loss": 0.7863,
+      "step": 13299
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019892161212776226,
+      "loss": 2.4067,
+      "step": 13300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019890040264267038,
+      "loss": 0.42,
+      "step": 13301
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000198878987765002,
+      "loss": 4.3805,
+      "step": 13302
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019885736753923083,
+      "loss": 2.5351,
+      "step": 13303
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019883554201025682,
+      "loss": 4.5302,
+      "step": 13304
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019881351122340654,
+      "loss": 5.1642,
+      "step": 13305
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019879127522443252,
+      "loss": 2.398,
+      "step": 13306
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019876883405951377,
+      "loss": 0.4271,
+      "step": 13307
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001987461877752552,
+      "loss": 2.5485,
+      "step": 13308
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019872333641868768,
+      "loss": 1.8853,
+      "step": 13309
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019870028003726815,
+      "loss": 1.9295,
+      "step": 13310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019867701867887913,
+      "loss": 1.1007,
+      "step": 13311
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019865355239182892,
+      "loss": 1.407,
+      "step": 13312
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001986298812248515,
+      "loss": 0.8586,
+      "step": 13313
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019860600522710612,
+      "loss": 1.4629,
+      "step": 13314
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019858192444817765,
+      "loss": 1.4129,
+      "step": 13315
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.000198557638938076,
+      "loss": 1.1878,
+      "step": 13316
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019853314874723662,
+      "loss": 0.3381,
+      "step": 13317
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019850845392651955,
+      "loss": 0.5123,
+      "step": 13318
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019848355452721018,
+      "loss": 1.7838,
+      "step": 13319
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019845845060101856,
+      "loss": 1.5117,
+      "step": 13320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001984331422000797,
+      "loss": 2.2059,
+      "step": 13321
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019840762937695295,
+      "loss": 4.68,
+      "step": 13322
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019838191218462254,
+      "loss": 1.8385,
+      "step": 13323
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019835599067649678,
+      "loss": 4.0444,
+      "step": 13324
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001983298649064086,
+      "loss": 2.4606,
+      "step": 13325
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001983035349286149,
+      "loss": 2.0967,
+      "step": 13326
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019827700079779682,
+      "loss": 1.2906,
+      "step": 13327
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019825026256905948,
+      "loss": 1.2672,
+      "step": 13328
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019822332029793168,
+      "loss": 2.892,
+      "step": 13329
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001981961740403663,
+      "loss": 0.3942,
+      "step": 13330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001981688238527394,
+      "loss": 2.7832,
+      "step": 13331
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019814126979185104,
+      "loss": 2.8726,
+      "step": 13332
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019811351191492427,
+      "loss": 3.7751,
+      "step": 13333
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019808555027960568,
+      "loss": 0.3548,
+      "step": 13334
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019805738494396485,
+      "loss": 3.5348,
+      "step": 13335
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001980290159664946,
+      "loss": 0.2286,
+      "step": 13336
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001980004434061104,
+      "loss": 2.2814,
+      "step": 13337
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019797166732215076,
+      "loss": 1.194,
+      "step": 13338
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001979426877743767,
+      "loss": 1.8035,
+      "step": 13339
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019791350482297178,
+      "loss": 0.3958,
+      "step": 13340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019788411852854213,
+      "loss": 4.2752,
+      "step": 13341
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001978545289521161,
+      "loss": 2.0977,
+      "step": 13342
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019782473615514404,
+      "loss": 0.8363,
+      "step": 13343
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001977947401994987,
+      "loss": 0.8689,
+      "step": 13344
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001977645411474743,
+      "loss": 1.9372,
+      "step": 13345
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019773413906178731,
+      "loss": 1.6429,
+      "step": 13346
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019770353400557546,
+      "loss": 2.6171,
+      "step": 13347
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001976727260423983,
+      "loss": 2.7898,
+      "step": 13348
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019764171523623642,
+      "loss": 1.9347,
+      "step": 13349
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001976105016514921,
+      "loss": 5.8243,
+      "step": 13350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019757908535298838,
+      "loss": 2.7342,
+      "step": 13351
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001975474664059695,
+      "loss": 1.4457,
+      "step": 13352
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001975156448761004,
+      "loss": 1.723,
+      "step": 13353
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019748362082946694,
+      "loss": 1.6231,
+      "step": 13354
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019745139433257526,
+      "loss": 2.0991,
+      "step": 13355
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019741896545235235,
+      "loss": 2.1386,
+      "step": 13356
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019738633425614503,
+      "loss": 2.5075,
+      "step": 13357
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019735350081172073,
+      "loss": 0.7568,
+      "step": 13358
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001973204651872665,
+      "loss": 2.2525,
+      "step": 13359
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019728722745138957,
+      "loss": 2.799,
+      "step": 13360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019725378767311685,
+      "loss": 3.0917,
+      "step": 13361
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019722014592189467,
+      "loss": 1.4527,
+      "step": 13362
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019718630226758907,
+      "loss": 4.1118,
+      "step": 13363
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001971522567804851,
+      "loss": 2.1055,
+      "step": 13364
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001971180095312873,
+      "loss": 2.8817,
+      "step": 13365
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019708356059111885,
+      "loss": 3.9021,
+      "step": 13366
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019704891003152219,
+      "loss": 2.6612,
+      "step": 13367
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001970140579244581,
+      "loss": 3.7433,
+      "step": 13368
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019697900434230629,
+      "loss": 1.5524,
+      "step": 13369
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019694374935786462,
+      "loss": 2.392,
+      "step": 13370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019690829304434916,
+      "loss": 0.6207,
+      "step": 13371
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019687263547539443,
+      "loss": 1.4797,
+      "step": 13372
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019683677672505253,
+      "loss": 1.0859,
+      "step": 13373
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001968007168677937,
+      "loss": 2.491,
+      "step": 13374
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019676445597850565,
+      "loss": 0.5522,
+      "step": 13375
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019672799413249342,
+      "loss": 1.5119,
+      "step": 13376
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019669133140547986,
+      "loss": 2.5741,
+      "step": 13377
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019665446787360446,
+      "loss": 3.0282,
+      "step": 13378
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001966174036134242,
+      "loss": 1.7048,
+      "step": 13379
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019658013870191258,
+      "loss": 1.8328,
+      "step": 13380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019654267321646005,
+      "loss": 3.086,
+      "step": 13381
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001965050072348734,
+      "loss": 5.4549,
+      "step": 13382
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.000196467140835376,
+      "loss": 2.1743,
+      "step": 13383
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001964290740966072,
+      "loss": 2.8126,
+      "step": 13384
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019639080709762277,
+      "loss": 3.1943,
+      "step": 13385
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001963523399178939,
+      "loss": 3.3105,
+      "step": 13386
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019631367263730796,
+      "loss": 2.2713,
+      "step": 13387
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019627480533616752,
+      "loss": 0.7037,
+      "step": 13388
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019623573809519088,
+      "loss": 1.137,
+      "step": 13389
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019619647099551116,
+      "loss": 3.5138,
+      "step": 13390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019615700411867694,
+      "loss": 1.3687,
+      "step": 13391
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001961173375466516,
+      "loss": 1.9809,
+      "step": 13392
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019607747136181293,
+      "loss": 5.0818,
+      "step": 13393
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019603740564695375,
+      "loss": 1.2934,
+      "step": 13394
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019599714048528072,
+      "loss": 0.4107,
+      "step": 13395
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001959566759604153,
+      "loss": 3.8018,
+      "step": 13396
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019591601215639237,
+      "loss": 1.7622,
+      "step": 13397
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019587514915766124,
+      "loss": 1.7229,
+      "step": 13398
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019583408704908437,
+      "loss": 3.397,
+      "step": 13399
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001957928259159382,
+      "loss": 2.7273,
+      "step": 13400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019575136584391208,
+      "loss": 3.1165,
+      "step": 13401
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001957097069191089,
+      "loss": 2.0304,
+      "step": 13402
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001956678492280442,
+      "loss": 1.8739,
+      "step": 13403
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019562579285764637,
+      "loss": 2.4187,
+      "step": 13404
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019558353789525667,
+      "loss": 2.2225,
+      "step": 13405
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019554108442862845,
+      "loss": 2.9022,
+      "step": 13406
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019549843254592738,
+      "loss": 3.5446,
+      "step": 13407
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019545558233573145,
+      "loss": 2.0161,
+      "step": 13408
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019541253388703016,
+      "loss": 2.5306,
+      "step": 13409
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019536928728922502,
+      "loss": 0.5674,
+      "step": 13410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001953258426321288,
+      "loss": 1.8642,
+      "step": 13411
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019528220000596585,
+      "loss": 5.2449,
+      "step": 13412
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001952383595013713,
+      "loss": 1.1161,
+      "step": 13413
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019519432120939167,
+      "loss": 1.8963,
+      "step": 13414
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019515008522148376,
+      "loss": 4.5337,
+      "step": 13415
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019510565162951545,
+      "loss": 0.5397,
+      "step": 13416
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019506102052576445,
+      "loss": 2.2529,
+      "step": 13417
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019501619200291916,
+      "loss": 2.5764,
+      "step": 13418
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019497116615407753,
+      "loss": 1.1773,
+      "step": 13419
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019492594307274775,
+      "loss": 2.9791,
+      "step": 13420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019488052285284718,
+      "loss": 2.1017,
+      "step": 13421
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019483490558870297,
+      "loss": 3.659,
+      "step": 13422
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019478909137505112,
+      "loss": 4.5187,
+      "step": 13423
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019474308030703697,
+      "loss": 1.6514,
+      "step": 13424
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019469687248021463,
+      "loss": 4.5724,
+      "step": 13425
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001946504679905465,
+      "loss": 2.2842,
+      "step": 13426
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019460386693440394,
+      "loss": 3.8901,
+      "step": 13427
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019455706940856599,
+      "loss": 1.9146,
+      "step": 13428
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019451007551022013,
+      "loss": 1.8924,
+      "step": 13429
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019451007551022013,
+      "loss": 4.7857,
+      "step": 13430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019446288533696136,
+      "loss": 2.2908,
+      "step": 13431
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019441549898679261,
+      "loss": 1.7642,
+      "step": 13432
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001943679165581238,
+      "loss": 5.1534,
+      "step": 13433
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019432013814977257,
+      "loss": 1.2053,
+      "step": 13434
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019427216386096315,
+      "loss": 2.008,
+      "step": 13435
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019422399379132662,
+      "loss": 3.1081,
+      "step": 13436
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019417562804090088,
+      "loss": 2.9389,
+      "step": 13437
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001941270667101301,
+      "loss": 1.7906,
+      "step": 13438
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019407830989986432,
+      "loss": 1.2962,
+      "step": 13439
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001940293577113602,
+      "loss": 1.1313,
+      "step": 13440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019398021024627946,
+      "loss": 3.2653,
+      "step": 13441
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019393086760668992,
+      "loss": 0.9637,
+      "step": 13442
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019388132989506428,
+      "loss": 0.57,
+      "step": 13443
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019383159721428076,
+      "loss": 1.2906,
+      "step": 13444
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001937816696676221,
+      "loss": 2.8313,
+      "step": 13445
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019373154735877613,
+      "loss": 3.2047,
+      "step": 13446
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001936812303918347,
+      "loss": 0.2856,
+      "step": 13447
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019363071887129442,
+      "loss": 4.3796,
+      "step": 13448
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019358001290205548,
+      "loss": 1.327,
+      "step": 13449
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019352911258942225,
+      "loss": 1.4732,
+      "step": 13450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001934780180391024,
+      "loss": 2.5133,
+      "step": 13451
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019342672935720732,
+      "loss": 1.3109,
+      "step": 13452
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019337524665025116,
+      "loss": 4.4628,
+      "step": 13453
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019332357002515153,
+      "loss": 0.484,
+      "step": 13454
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019327169958922814,
+      "loss": 2.1059,
+      "step": 13455
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019321963545020375,
+      "loss": 0.94,
+      "step": 13456
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019316737771620318,
+      "loss": 2.216,
+      "step": 13457
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019311492649575315,
+      "loss": 2.2293,
+      "step": 13458
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019306228189778254,
+      "loss": 1.1256,
+      "step": 13459
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019300944403162143,
+      "loss": 5.0566,
+      "step": 13460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001929564130070017,
+      "loss": 0.4127,
+      "step": 13461
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019290318893405592,
+      "loss": 2.757,
+      "step": 13462
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019284977192331805,
+      "loss": 1.5197,
+      "step": 13463
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019279616208572225,
+      "loss": 2.7636,
+      "step": 13464
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019274235953260365,
+      "loss": 2.0189,
+      "step": 13465
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001926883643756972,
+      "loss": 1.0827,
+      "step": 13466
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019263417672713782,
+      "loss": 2.3882,
+      "step": 13467
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019257979669946065,
+      "loss": 4.6568,
+      "step": 13468
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001925252244055999,
+      "loss": 0.9684,
+      "step": 13469
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019247045995888907,
+      "loss": 1.993,
+      "step": 13470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001924155034730612,
+      "loss": 1.6783,
+      "step": 13471
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001923603550622475,
+      "loss": 3.4215,
+      "step": 13472
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019230501484097836,
+      "loss": 0.7344,
+      "step": 13473
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019224948292418198,
+      "loss": 2.9983,
+      "step": 13474
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019219375942718518,
+      "loss": 0.3013,
+      "step": 13475
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001921378444657121,
+      "loss": 2.07,
+      "step": 13476
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019208173815588514,
+      "loss": 0.84,
+      "step": 13477
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019202544061422338,
+      "loss": 1.8629,
+      "step": 13478
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019196895195764372,
+      "loss": 1.1796,
+      "step": 13479
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019191227230345935,
+      "loss": 1.7684,
+      "step": 13480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019185540176938073,
+      "loss": 2.6012,
+      "step": 13481
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019179834047351407,
+      "loss": 4.6916,
+      "step": 13482
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019174108853436242,
+      "loss": 2.8728,
+      "step": 13483
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001916836460708241,
+      "loss": 2.9915,
+      "step": 13484
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019162601320219373,
+      "loss": 2.4744,
+      "step": 13485
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019156819004816078,
+      "loss": 4.8161,
+      "step": 13486
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019151017672881037,
+      "loss": 1.0267,
+      "step": 13487
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019145197336462217,
+      "loss": 2.6746,
+      "step": 13488
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001913935800764708,
+      "loss": 2.638,
+      "step": 13489
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001913349969856253,
+      "loss": 1.0496,
+      "step": 13490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019127622421374858,
+      "loss": 0.5388,
+      "step": 13491
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019121726188289792,
+      "loss": 2.5415,
+      "step": 13492
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001911581101155238,
+      "loss": 4.021,
+      "step": 13493
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019109876903447063,
+      "loss": 1.2218,
+      "step": 13494
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019103923876297537,
+      "loss": 2.3129,
+      "step": 13495
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019097951942466854,
+      "loss": 2.3934,
+      "step": 13496
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001909196111435728,
+      "loss": 1.8101,
+      "step": 13497
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019085951404410325,
+      "loss": 1.942,
+      "step": 13498
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019079922825106756,
+      "loss": 3.4929,
+      "step": 13499
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019073875388966465,
+      "loss": 1.1042,
+      "step": 13500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019067809108548565,
+      "loss": 2.6918,
+      "step": 13501
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001906172399645126,
+      "loss": 5.4233,
+      "step": 13502
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019055620065311867,
+      "loss": 4.2893,
+      "step": 13503
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019049497327806826,
+      "loss": 0.7414,
+      "step": 13504
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001904335579665157,
+      "loss": 0.3335,
+      "step": 13505
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001903719548460063,
+      "loss": 1.237,
+      "step": 13506
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019031016404447474,
+      "loss": 3.3326,
+      "step": 13507
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019024818569024615,
+      "loss": 2.3395,
+      "step": 13508
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019018601991203444,
+      "loss": 0.8063,
+      "step": 13509
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019012366683894346,
+      "loss": 2.5127,
+      "step": 13510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019006112660046535,
+      "loss": 1.823,
+      "step": 13511
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018999839932648164,
+      "loss": 2.4239,
+      "step": 13512
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018993548514726163,
+      "loss": 3.6926,
+      "step": 13513
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018987238419346337,
+      "loss": 2.6007,
+      "step": 13514
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001898090965961322,
+      "loss": 3.7073,
+      "step": 13515
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018974562248670168,
+      "loss": 1.8196,
+      "step": 13516
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001896819619969921,
+      "loss": 1.5341,
+      "step": 13517
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001896181152592115,
+      "loss": 2.1704,
+      "step": 13518
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018955408240595395,
+      "loss": 2.7689,
+      "step": 13519
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018948986357020066,
+      "loss": 2.0454,
+      "step": 13520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018942545888531897,
+      "loss": 1.0469,
+      "step": 13521
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018936086848506177,
+      "loss": 4.305,
+      "step": 13522
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018929609250356827,
+      "loss": 3.1146,
+      "step": 13523
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018923113107536241,
+      "loss": 2.3911,
+      "step": 13524
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018916598433535388,
+      "loss": 3.5733,
+      "step": 13525
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001891006524188367,
+      "loss": 3.3646,
+      "step": 13526
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018903513546148997,
+      "loss": 4.0194,
+      "step": 13527
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018896943359937652,
+      "loss": 1.6467,
+      "step": 13528
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018890354696894375,
+      "loss": 3.7406,
+      "step": 13529
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018883747570702233,
+      "loss": 2.5149,
+      "step": 13530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001887712199508264,
+      "loss": 3.0768,
+      "step": 13531
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018870477983795367,
+      "loss": 0.6776,
+      "step": 13532
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001886381555063843,
+      "loss": 1.5509,
+      "step": 13533
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018857134709448096,
+      "loss": 3.9704,
+      "step": 13534
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018850435474098917,
+      "loss": 0.8805,
+      "step": 13535
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018843717858503577,
+      "loss": 1.3321,
+      "step": 13536
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001883698187661299,
+      "loss": 0.4598,
+      "step": 13537
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001883022754241616,
+      "loss": 2.1271,
+      "step": 13538
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018823454869940255,
+      "loss": 1.1383,
+      "step": 13539
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018816663873250477,
+      "loss": 3.9948,
+      "step": 13540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001880985456645014,
+      "loss": 1.9956,
+      "step": 13541
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018803026963680513,
+      "loss": 3.409,
+      "step": 13542
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018796181079120942,
+      "loss": 2.173,
+      "step": 13543
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018789316926988658,
+      "loss": 0.7641,
+      "step": 13544
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018782434521538908,
+      "loss": 1.5543,
+      "step": 13545
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018775533877064774,
+      "loss": 2.3389,
+      "step": 13546
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018768615007897283,
+      "loss": 0.5759,
+      "step": 13547
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018761677928405245,
+      "loss": 3.2902,
+      "step": 13548
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018754722652995355,
+      "loss": 3.215,
+      "step": 13549
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018747749196112033,
+      "loss": 2.493,
+      "step": 13550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018740757572237519,
+      "loss": 0.1831,
+      "step": 13551
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001873374779589171,
+      "loss": 3.5871,
+      "step": 13552
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001872671988163227,
+      "loss": 0.3624,
+      "step": 13553
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001871967384405451,
+      "loss": 2.2928,
+      "step": 13554
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018712609697791344,
+      "loss": 3.4478,
+      "step": 13555
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001870552745751335,
+      "loss": 1.3842,
+      "step": 13556
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018698427137928627,
+      "loss": 2.737,
+      "step": 13557
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018691308753782874,
+      "loss": 4.2301,
+      "step": 13558
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018684172319859246,
+      "loss": 1.9262,
+      "step": 13559
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018677017850978457,
+      "loss": 3.5285,
+      "step": 13560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018669845361998614,
+      "loss": 4.1565,
+      "step": 13561
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018662654867815246,
+      "loss": 3.135,
+      "step": 13562
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018655446383361338,
+      "loss": 1.7509,
+      "step": 13563
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001864821992360717,
+      "loss": 1.2766,
+      "step": 13564
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018640975503560368,
+      "loss": 2.7312,
+      "step": 13565
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018633713138265894,
+      "loss": 2.5683,
+      "step": 13566
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018626432842805922,
+      "loss": 1.8332,
+      "step": 13567
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018619134632299927,
+      "loss": 0.7519,
+      "step": 13568
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018611818521904513,
+      "loss": 5.2311,
+      "step": 13569
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018604484526813536,
+      "loss": 3.8113,
+      "step": 13570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018597132662257924,
+      "loss": 0.8749,
+      "step": 13571
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018589762943505779,
+      "loss": 2.6086,
+      "step": 13572
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018582375385862217,
+      "loss": 1.6448,
+      "step": 13573
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001857497000466948,
+      "loss": 0.7769,
+      "step": 13574
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018567546815306728,
+      "loss": 0.8433,
+      "step": 13575
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018560105833190205,
+      "loss": 3.1232,
+      "step": 13576
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018552647073773022,
+      "loss": 0.5557,
+      "step": 13577
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018545170552545278,
+      "loss": 2.8561,
+      "step": 13578
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001853767628503389,
+      "loss": 3.3536,
+      "step": 13579
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018530164286802703,
+      "loss": 1.1938,
+      "step": 13580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018522634573452313,
+      "loss": 2.2356,
+      "step": 13581
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018515087160620177,
+      "loss": 0.3993,
+      "step": 13582
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018507522063980435,
+      "loss": 1.5022,
+      "step": 13583
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018499939299244017,
+      "loss": 1.2913,
+      "step": 13584
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018492338882158523,
+      "loss": 1.4518,
+      "step": 13585
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018484720828508184,
+      "loss": 0.6645,
+      "step": 13586
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018477085154113912,
+      "loss": 3.2729,
+      "step": 13587
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018469431874833155,
+      "loss": 1.4796,
+      "step": 13588
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001846176100655998,
+      "loss": 3.3716,
+      "step": 13589
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001845407256522492,
+      "loss": 3.016,
+      "step": 13590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018446366566795072,
+      "loss": 2.7548,
+      "step": 13591
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018438643027273943,
+      "loss": 3.0761,
+      "step": 13592
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001843090196270146,
+      "loss": 1.0617,
+      "step": 13593
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018423143389154016,
+      "loss": 5.568,
+      "step": 13594
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018415367322744276,
+      "loss": 1.1607,
+      "step": 13595
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018407573779621316,
+      "loss": 3.2203,
+      "step": 13596
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018399762775970452,
+      "loss": 1.3107,
+      "step": 13597
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001839193432801326,
+      "loss": 2.5113,
+      "step": 13598
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018384088452007592,
+      "loss": 0.4414,
+      "step": 13599
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018376225164247432,
+      "loss": 0.4975,
+      "step": 13600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018368344481062992,
+      "loss": 2.3423,
+      "step": 13601
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018360446418820537,
+      "loss": 3.2007,
+      "step": 13602
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018352530993922498,
+      "loss": 1.5935,
+      "step": 13603
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018344598222807287,
+      "loss": 1.7615,
+      "step": 13604
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018336648121949418,
+      "loss": 1.7681,
+      "step": 13605
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018328680707859323,
+      "loss": 3.5121,
+      "step": 13606
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018320695997083456,
+      "loss": 2.9808,
+      "step": 13607
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018312694006204124,
+      "loss": 3.9643,
+      "step": 13608
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018304674751839595,
+      "loss": 1.6548,
+      "step": 13609
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018296638250643905,
+      "loss": 0.4704,
+      "step": 13610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018288584519307,
+      "loss": 1.188,
+      "step": 13611
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018280513574554517,
+      "loss": 2.8573,
+      "step": 13612
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018272425433147924,
+      "loss": 0.8213,
+      "step": 13613
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001826432011188433,
+      "loss": 1.4166,
+      "step": 13614
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018256197627596591,
+      "loss": 1.5117,
+      "step": 13615
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018248057997153136,
+      "loss": 3.5412,
+      "step": 13616
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018239901237458055,
+      "loss": 1.6022,
+      "step": 13617
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018231727365451017,
+      "loss": 1.2003,
+      "step": 13618
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018223536398107165,
+      "loss": 1.2294,
+      "step": 13619
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001821532835243723,
+      "loss": 2.6967,
+      "step": 13620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018207103245487325,
+      "loss": 2.7798,
+      "step": 13621
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018198861094339077,
+      "loss": 1.2322,
+      "step": 13622
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018190601916109452,
+      "loss": 1.5981,
+      "step": 13623
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018182325727950777,
+      "loss": 0.4181,
+      "step": 13624
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018174032547050762,
+      "loss": 2.1182,
+      "step": 13625
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018165722390632328,
+      "loss": 3.0314,
+      "step": 13626
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018157395275953729,
+      "loss": 1.5932,
+      "step": 13627
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001814905122030838,
+      "loss": 2.6192,
+      "step": 13628
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018140690241024877,
+      "loss": 1.7937,
+      "step": 13629
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018132312355467024,
+      "loss": 2.7557,
+      "step": 13630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018123917581033648,
+      "loss": 2.4332,
+      "step": 13631
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018115505935158738,
+      "loss": 2.9159,
+      "step": 13632
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018107077435311245,
+      "loss": 3.036,
+      "step": 13633
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018098632098995188,
+      "loss": 2.752,
+      "step": 13634
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001809016994374948,
+      "loss": 1.0756,
+      "step": 13635
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018081690987148044,
+      "loss": 0.9226,
+      "step": 13636
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001807319524679961,
+      "loss": 1.457,
+      "step": 13637
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001806468274034785,
+      "loss": 2.5139,
+      "step": 13638
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001805615348547117,
+      "loss": 1.1331,
+      "step": 13639
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018047607499882838,
+      "loss": 1.674,
+      "step": 13640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018039044801330798,
+      "loss": 1.4363,
+      "step": 13641
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018030465407597768,
+      "loss": 1.211,
+      "step": 13642
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018021869336501064,
+      "loss": 2.1563,
+      "step": 13643
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018013256605892727,
+      "loss": 2.6514,
+      "step": 13644
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000180046272336593,
+      "loss": 1.1384,
+      "step": 13645
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017995981237721974,
+      "loss": 2.3814,
+      "step": 13646
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017987318636036382,
+      "loss": 2.0756,
+      "step": 13647
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017978639446592708,
+      "loss": 1.2032,
+      "step": 13648
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017969943687415576,
+      "loss": 1.8965,
+      "step": 13649
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017961231376563972,
+      "loss": 0.4256,
+      "step": 13650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017952502532131326,
+      "loss": 1.3411,
+      "step": 13651
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017943757172245327,
+      "loss": 2.5408,
+      "step": 13652
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017934995315068051,
+      "loss": 2.2316,
+      "step": 13653
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017926216978795737,
+      "loss": 2.515,
+      "step": 13654
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017917422181658953,
+      "loss": 2.6714,
+      "step": 13655
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017908610941922367,
+      "loss": 1.9514,
+      "step": 13656
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017899783277884805,
+      "loss": 1.1828,
+      "step": 13657
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017890939207879268,
+      "loss": 1.1962,
+      "step": 13658
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001788207875027276,
+      "loss": 1.9756,
+      "step": 13659
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017873201923466317,
+      "loss": 3.8794,
+      "step": 13660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017864308745895043,
+      "loss": 3.6483,
+      "step": 13661
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000178553992360279,
+      "loss": 1.3792,
+      "step": 13662
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000178553992360279,
+      "loss": 4.707,
+      "step": 13663
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001784647341236786,
+      "loss": 2.1319,
+      "step": 13664
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017837531293451694,
+      "loss": 1.0547,
+      "step": 13665
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017828572897850091,
+      "loss": 3.0173,
+      "step": 13666
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017819598244167467,
+      "loss": 4.4241,
+      "step": 13667
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017810607351042076,
+      "loss": 2.6076,
+      "step": 13668
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017801600237145825,
+      "loss": 1.7473,
+      "step": 13669
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017792576921184385,
+      "loss": 1.7035,
+      "step": 13670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017783537421896988,
+      "loss": 3.6926,
+      "step": 13671
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001777448175805657,
+      "loss": 2.413,
+      "step": 13672
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00017765409948469535,
+      "loss": 2.9918,
+      "step": 13673
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001775632201197592,
+      "loss": 1.1439,
+      "step": 13674
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001774721796744916,
+      "loss": 2.8001,
+      "step": 13675
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001773809783379623,
+      "loss": 2.4433,
+      "step": 13676
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001772896162995743,
+      "loss": 1.882,
+      "step": 13677
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017719809374906525,
+      "loss": 2.5811,
+      "step": 13678
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017710641087650522,
+      "loss": 1.3499,
+      "step": 13679
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017701456787229817,
+      "loss": 3.2012,
+      "step": 13680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017692256492717967,
+      "loss": 2.7788,
+      "step": 13681
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017683040223221815,
+      "loss": 1.4575,
+      "step": 13682
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017673807997881382,
+      "loss": 0.3916,
+      "step": 13683
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001766455983586975,
+      "loss": 1.0767,
+      "step": 13684
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000176552957563932,
+      "loss": 1.1194,
+      "step": 13685
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001764601577869097,
+      "loss": 2.9618,
+      "step": 13686
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017636719922035408,
+      "loss": 0.8884,
+      "step": 13687
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001762740820573177,
+      "loss": 1.7566,
+      "step": 13688
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017618080649118252,
+      "loss": 2.164,
+      "step": 13689
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017608737271566013,
+      "loss": 2.4872,
+      "step": 13690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017599378092479002,
+      "loss": 1.2983,
+      "step": 13691
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017590003131293976,
+      "loss": 0.955,
+      "step": 13692
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017580612407480542,
+      "loss": 1.7945,
+      "step": 13693
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017571205940540953,
+      "loss": 3.9856,
+      "step": 13694
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017561783750010248,
+      "loss": 2.0996,
+      "step": 13695
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001755234585545602,
+      "loss": 4.5244,
+      "step": 13696
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001754289227647858,
+      "loss": 2.8657,
+      "step": 13697
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017533423032710713,
+      "loss": 2.8865,
+      "step": 13698
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017523938143817824,
+      "loss": 0.4349,
+      "step": 13699
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017514437629497724,
+      "loss": 2.883,
+      "step": 13700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017504921509480767,
+      "loss": 1.023,
+      "step": 13701
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017495389803529618,
+      "loss": 1.1037,
+      "step": 13702
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017485842531439404,
+      "loss": 2.2192,
+      "step": 13703
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017476279713037496,
+      "loss": 1.7248,
+      "step": 13704
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017466701368183628,
+      "loss": 1.0656,
+      "step": 13705
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017457107516769696,
+      "loss": 1.0666,
+      "step": 13706
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017447498178719894,
+      "loss": 1.2213,
+      "step": 13707
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017437873373990476,
+      "loss": 1.1894,
+      "step": 13708
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017428233122569924,
+      "loss": 1.4157,
+      "step": 13709
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017418577444478688,
+      "loss": 3.62,
+      "step": 13710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017408906359769365,
+      "loss": 4.3905,
+      "step": 13711
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017399219888526436,
+      "loss": 0.2779,
+      "step": 13712
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017389518050866433,
+      "loss": 3.6862,
+      "step": 13713
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017379800866937773,
+      "loss": 1.979,
+      "step": 13714
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017370068356920696,
+      "loss": 0.5876,
+      "step": 13715
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001736032054102734,
+      "loss": 3.1716,
+      "step": 13716
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017350557439501561,
+      "loss": 1.8929,
+      "step": 13717
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001734077907261904,
+      "loss": 4.613,
+      "step": 13718
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001733098546068709,
+      "loss": 1.5211,
+      "step": 13719
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017321176624044684,
+      "loss": 2.0658,
+      "step": 13720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001731135258306249,
+      "loss": 2.6633,
+      "step": 13721
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017301513358142657,
+      "loss": 1.7015,
+      "step": 13722
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001729165896971895,
+      "loss": 1.7692,
+      "step": 13723
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017281789438256567,
+      "loss": 2.4246,
+      "step": 13724
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001727190478425215,
+      "loss": 2.3191,
+      "step": 13725
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001726200502823382,
+      "loss": 2.9762,
+      "step": 13726
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001725209019076096,
+      "loss": 1.9118,
+      "step": 13727
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001724216029242438,
+      "loss": 0.334,
+      "step": 13728
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017232215353846057,
+      "loss": 0.3255,
+      "step": 13729
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017222255395679315,
+      "loss": 2.2811,
+      "step": 13730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001721228043860856,
+      "loss": 2.1549,
+      "step": 13731
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017202290503349455,
+      "loss": 0.3171,
+      "step": 13732
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017192285610648664,
+      "loss": 2.7435,
+      "step": 13733
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017182265781284014,
+      "loss": 1.837,
+      "step": 13734
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001717223103606426,
+      "loss": 3.5658,
+      "step": 13735
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017162181395829218,
+      "loss": 1.2473,
+      "step": 13736
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017152116881449556,
+      "loss": 1.1336,
+      "step": 13737
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017142037513826917,
+      "loss": 1.6935,
+      "step": 13738
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017131943313893698,
+      "loss": 2.0888,
+      "step": 13739
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000171218343026132,
+      "loss": 3.3063,
+      "step": 13740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001711171050097938,
+      "loss": 4.0159,
+      "step": 13741
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001710157193001701,
+      "loss": 1.5212,
+      "step": 13742
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001709141861078144,
+      "loss": 1.6429,
+      "step": 13743
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001708125056435874,
+      "loss": 0.6553,
+      "step": 13744
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017071067811865476,
+      "loss": 0.659,
+      "step": 13745
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001706087037444882,
+      "loss": 2.1401,
+      "step": 13746
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001705065827328645,
+      "loss": 0.4529,
+      "step": 13747
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017040431529586413,
+      "loss": 1.1278,
+      "step": 13748
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017030190164587272,
+      "loss": 1.6284,
+      "step": 13749
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017019934199557853,
+      "loss": 0.6611,
+      "step": 13750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00017009663655797395,
+      "loss": 0.5289,
+      "step": 13751
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001699937855463536,
+      "loss": 0.5692,
+      "step": 13752
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016989078917431413,
+      "loss": 2.0685,
+      "step": 13753
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016978764765575508,
+      "loss": 1.0502,
+      "step": 13754
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001696843612048765,
+      "loss": 2.2371,
+      "step": 13755
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001695809300361795,
+      "loss": 0.5467,
+      "step": 13756
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016947735436446653,
+      "loss": 2.1355,
+      "step": 13757
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016937363440483902,
+      "loss": 1.4358,
+      "step": 13758
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016926977037269914,
+      "loss": 2.9594,
+      "step": 13759
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016916576248374727,
+      "loss": 0.7146,
+      "step": 13760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016906161095398348,
+      "loss": 2.4755,
+      "step": 13761
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016895731599970518,
+      "loss": 2.0548,
+      "step": 13762
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016885287783750873,
+      "loss": 0.9362,
+      "step": 13763
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016874829668428668,
+      "loss": 1.3933,
+      "step": 13764
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001686435727572298,
+      "loss": 2.3946,
+      "step": 13765
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016853870627382418,
+      "loss": 0.9386,
+      "step": 13766
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016843369745185303,
+      "loss": 2.3165,
+      "step": 13767
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016832854650939414,
+      "loss": 1.7201,
+      "step": 13768
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016822325366482147,
+      "loss": 2.5374,
+      "step": 13769
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016811781913680273,
+      "loss": 0.8053,
+      "step": 13770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016801224314430074,
+      "loss": 3.4363,
+      "step": 13771
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016790652590657128,
+      "loss": 2.2152,
+      "step": 13772
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016780066764316426,
+      "loss": 1.6343,
+      "step": 13773
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016769466857392167,
+      "loss": 1.1569,
+      "step": 13774
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016758852891897877,
+      "loss": 0.3433,
+      "step": 13775
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016748224889876186,
+      "loss": 1.6807,
+      "step": 13776
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016737582873398973,
+      "loss": 3.1625,
+      "step": 13777
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001672692686456712,
+      "loss": 3.9938,
+      "step": 13778
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001671625688551064,
+      "loss": 1.8633,
+      "step": 13779
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001670557295838857,
+      "loss": 3.9866,
+      "step": 13780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016694875105388829,
+      "loss": 1.6183,
+      "step": 13781
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016684163348728358,
+      "loss": 0.1582,
+      "step": 13782
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000166734377106529,
+      "loss": 0.3083,
+      "step": 13783
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016662698213437024,
+      "loss": 2.3691,
+      "step": 13784
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016651944879384156,
+      "loss": 3.5062,
+      "step": 13785
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001664117773082639,
+      "loss": 3.6245,
+      "step": 13786
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016630396790124496,
+      "loss": 2.7309,
+      "step": 13787
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016619602079667977,
+      "loss": 1.5801,
+      "step": 13788
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016608793621874818,
+      "loss": 2.497,
+      "step": 13789
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016597971439191665,
+      "loss": 2.6441,
+      "step": 13790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016587135554093564,
+      "loss": 2.4552,
+      "step": 13791
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001657628598908412,
+      "loss": 2.3082,
+      "step": 13792
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016565422766695234,
+      "loss": 4.1604,
+      "step": 13793
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016554545909487295,
+      "loss": 2.8186,
+      "step": 13794
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016543655440048882,
+      "loss": 1.4229,
+      "step": 13795
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016532751380996964,
+      "loss": 1.2163,
+      "step": 13796
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016521833754976618,
+      "loss": 0.8685,
+      "step": 13797
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001651090258466121,
+      "loss": 2.63,
+      "step": 13798
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001649995789275212,
+      "loss": 2.4305,
+      "step": 13799
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001648899970197892,
+      "loss": 2.9635,
+      "step": 13800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016478028035099107,
+      "loss": 2.7123,
+      "step": 13801
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016467042914898277,
+      "loss": 2.0914,
+      "step": 13802
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001645604436418985,
+      "loss": 0.6906,
+      "step": 13803
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016445032405815253,
+      "loss": 3.0206,
+      "step": 13804
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001643400706264364,
+      "loss": 2.3108,
+      "step": 13805
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016422968357572085,
+      "loss": 2.0051,
+      "step": 13806
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001641191631352529,
+      "loss": 1.8794,
+      "step": 13807
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016400850953455774,
+      "loss": 6.0771,
+      "step": 13808
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016389772300343596,
+      "loss": 2.2769,
+      "step": 13809
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016378680377196508,
+      "loss": 0.9042,
+      "step": 13810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016367575207049822,
+      "loss": 0.3329,
+      "step": 13811
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001635645681296627,
+      "loss": 3.5748,
+      "step": 13812
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016345325218036173,
+      "loss": 3.8443,
+      "step": 13813
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016334180445377164,
+      "loss": 2.5181,
+      "step": 13814
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001632302251813426,
+      "loss": 2.4621,
+      "step": 13815
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001631185145947987,
+      "loss": 1.791,
+      "step": 13816
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016300667292613597,
+      "loss": 4.039,
+      "step": 13817
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016289470040762257,
+      "loss": 1.3487,
+      "step": 13818
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001627825972717994,
+      "loss": 0.8773,
+      "step": 13819
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016267036375147733,
+      "loss": 1.02,
+      "step": 13820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016255800007973923,
+      "loss": 2.0642,
+      "step": 13821
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016244550648993714,
+      "loss": 1.4018,
+      "step": 13822
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016233288321569397,
+      "loss": 2.7941,
+      "step": 13823
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001622201304909009,
+      "loss": 2.0248,
+      "step": 13824
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016210724854971904,
+      "loss": 2.6316,
+      "step": 13825
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016199423762657672,
+      "loss": 0.1993,
+      "step": 13826
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016188109795617128,
+      "loss": 0.8736,
+      "step": 13827
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001617678297734663,
+      "loss": 1.7861,
+      "step": 13828
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016165443331369334,
+      "loss": 2.6991,
+      "step": 13829
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016154090881234933,
+      "loss": 2.6623,
+      "step": 13830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001614272565051981,
+      "loss": 1.1015,
+      "step": 13831
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016131347662826792,
+      "loss": 3.112,
+      "step": 13832
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016119956941785306,
+      "loss": 0.7257,
+      "step": 13833
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016108553511051104,
+      "loss": 1.5338,
+      "step": 13834
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001609713739430646,
+      "loss": 1.9108,
+      "step": 13835
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016085708615259863,
+      "loss": 2.2081,
+      "step": 13836
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016074267197646225,
+      "loss": 3.1972,
+      "step": 13837
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016062813165226592,
+      "loss": 3.1346,
+      "step": 13838
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016051346541788312,
+      "loss": 5.0134,
+      "step": 13839
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016039867351144775,
+      "loss": 2.6294,
+      "step": 13840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016028375617135586,
+      "loss": 2.673,
+      "step": 13841
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00016016871363626277,
+      "loss": 3.7295,
+      "step": 13842
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001600535461450847,
+      "loss": 3.2673,
+      "step": 13843
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015993825393699742,
+      "loss": 1.7559,
+      "step": 13844
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015982283725143507,
+      "loss": 3.5116,
+      "step": 13845
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015970729632809038,
+      "loss": 4.5232,
+      "step": 13846
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015959163140691481,
+      "loss": 2.3127,
+      "step": 13847
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015947584272811636,
+      "loss": 1.7431,
+      "step": 13848
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000159359930532161,
+      "loss": 0.5049,
+      "step": 13849
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015924389505977057,
+      "loss": 1.6266,
+      "step": 13850
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015912773655192276,
+      "loss": 0.453,
+      "step": 13851
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015901145524985173,
+      "loss": 0.5529,
+      "step": 13852
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015889505139504543,
+      "loss": 3.2927,
+      "step": 13853
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001587785252292475,
+      "loss": 2.4793,
+      "step": 13854
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015866187699445453,
+      "loss": 1.9903,
+      "step": 13855
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001585451069329176,
+      "loss": 3.4106,
+      "step": 13856
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00015842821528713982,
+      "loss": 1.91,
+      "step": 13857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015831120229987781,
+      "loss": 2.0072,
+      "step": 13858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001581940682141391,
+      "loss": 4.1817,
+      "step": 13859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001580768132731839,
+      "loss": 1.2055,
+      "step": 13860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015795943772052215,
+      "loss": 4.2827,
+      "step": 13861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015784194179991552,
+      "loss": 2.9444,
+      "step": 13862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015772432575537448,
+      "loss": 0.7027,
+      "step": 13863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001576065898311601,
+      "loss": 1.1167,
+      "step": 13864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015748873427178127,
+      "loss": 1.1005,
+      "step": 13865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001573707593219965,
+      "loss": 0.6231,
+      "step": 13866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015725266522681111,
+      "loss": 1.773,
+      "step": 13867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015713445223147893,
+      "loss": 2.2046,
+      "step": 13868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001570161205814996,
+      "loss": 0.5759,
+      "step": 13869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015689767052262036,
+      "loss": 0.4679,
+      "step": 13870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015677910230083323,
+      "loss": 1.9127,
+      "step": 13871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015666041616237673,
+      "loss": 0.1707,
+      "step": 13872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015654161235373305,
+      "loss": 0.5952,
+      "step": 13873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015642269112162976,
+      "loss": 3.1948,
+      "step": 13874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015630365271303824,
+      "loss": 4.7467,
+      "step": 13875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001561844973751722,
+      "loss": 2.5689,
+      "step": 13876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015606522535548963,
+      "loss": 1.4684,
+      "step": 13877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015594583690168974,
+      "loss": 1.0388,
+      "step": 13878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015582633226171358,
+      "loss": 1.36,
+      "step": 13879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015570671168374446,
+      "loss": 2.5437,
+      "step": 13880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001555869754162055,
+      "loss": 1.6981,
+      "step": 13881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015546712370776016,
+      "loss": 2.3568,
+      "step": 13882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015534715680731238,
+      "loss": 3.5716,
+      "step": 13883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015522707496400435,
+      "loss": 3.0745,
+      "step": 13884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015510687842721797,
+      "loss": 2.376,
+      "step": 13885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015498656744657226,
+      "loss": 1.2644,
+      "step": 13886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000154866142271925,
+      "loss": 2.5826,
+      "step": 13887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001547456031533701,
+      "loss": 0.8802,
+      "step": 13888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015462495034123915,
+      "loss": 2.1502,
+      "step": 13889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001545041840860986,
+      "loss": 2.9584,
+      "step": 13890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015438330463875196,
+      "loss": 2.2422,
+      "step": 13891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015426231225023634,
+      "loss": 0.2839,
+      "step": 13892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015414120717182473,
+      "loss": 1.1424,
+      "step": 13893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015401998965502292,
+      "loss": 2.2933,
+      "step": 13894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015389865995157144,
+      "loss": 0.7637,
+      "step": 13895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001537772183134426,
+      "loss": 0.1351,
+      "step": 13896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015365566499284225,
+      "loss": 1.2403,
+      "step": 13897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001535340002422072,
+      "loss": 3.1208,
+      "step": 13898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001534122243142067,
+      "loss": 2.1755,
+      "step": 13899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015329033746173972,
+      "loss": 0.725,
+      "step": 13900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001531683399379369,
+      "loss": 3.2372,
+      "step": 13901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015304623199615736,
+      "loss": 1.2289,
+      "step": 13902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015292401388999083,
+      "loss": 3.6876,
+      "step": 13903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015280168587325456,
+      "loss": 3.8569,
+      "step": 13904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001526792481999954,
+      "loss": 1.4103,
+      "step": 13905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015255670112448652,
+      "loss": 1.4369,
+      "step": 13906
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015243404490122942,
+      "loss": 2.3812,
+      "step": 13907
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015231127978495204,
+      "loss": 2.3749,
+      "step": 13908
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015218840603060802,
+      "loss": 2.3466,
+      "step": 13909
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015206542389337662,
+      "loss": 4.1678,
+      "step": 13910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015194233362866302,
+      "loss": 3.7021,
+      "step": 13911
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015181913549209606,
+      "loss": 0.589,
+      "step": 13912
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001516958297395286,
+      "loss": 0.5945,
+      "step": 13913
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015157241662703798,
+      "loss": 1.4045,
+      "step": 13914
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015144889641092318,
+      "loss": 4.2165,
+      "step": 13915
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015132526934770693,
+      "loss": 1.3776,
+      "step": 13916
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015120153569413248,
+      "loss": 3.0818,
+      "step": 13917
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015107769570716587,
+      "loss": 1.204,
+      "step": 13918
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015095374964399255,
+      "loss": 2.7913,
+      "step": 13919
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015082969776201966,
+      "loss": 0.4508,
+      "step": 13920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015070554031887276,
+      "loss": 1.3981,
+      "step": 13921
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015058127757239787,
+      "loss": 2.2408,
+      "step": 13922
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015045690978065856,
+      "loss": 1.449,
+      "step": 13923
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015033243720193764,
+      "loss": 2.3293,
+      "step": 13924
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015020786009473447,
+      "loss": 0.9755,
+      "step": 13925
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00015008317871776657,
+      "loss": 0.4745,
+      "step": 13926
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001499583933299669,
+      "loss": 2.0732,
+      "step": 13927
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014983350419048557,
+      "loss": 1.586,
+      "step": 13928
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014970851155868698,
+      "loss": 1.6416,
+      "step": 13929
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014958341569415163,
+      "loss": 2.522,
+      "step": 13930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014945821685667326,
+      "loss": 3.9611,
+      "step": 13931
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001493329153062606,
+      "loss": 1.6368,
+      "step": 13932
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014920751130313454,
+      "loss": 1.1171,
+      "step": 13933
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001490820051077299,
+      "loss": 1.5981,
+      "step": 13934
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014895639698069257,
+      "loss": 1.7269,
+      "step": 13935
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014883068718288131,
+      "loss": 1.5581,
+      "step": 13936
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014870487597536483,
+      "loss": 3.4768,
+      "step": 13937
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014857896361942328,
+      "loss": 2.5388,
+      "step": 13938
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014845295037654703,
+      "loss": 1.5448,
+      "step": 13939
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001483268365084352,
+      "loss": 0.244,
+      "step": 13940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001482006222769959,
+      "loss": 1.5256,
+      "step": 13941
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014807430794434675,
+      "loss": 3.0928,
+      "step": 13942
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014794789377281186,
+      "loss": 3.4821,
+      "step": 13943
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014782138002492393,
+      "loss": 4.0341,
+      "step": 13944
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014769476696342162,
+      "loss": 0.6238,
+      "step": 13945
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014756805485124976,
+      "loss": 2.2892,
+      "step": 13946
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014744124395155985,
+      "loss": 1.4819,
+      "step": 13947
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001473143345277073,
+      "loss": 1.2263,
+      "step": 13948
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014718732684325335,
+      "loss": 0.6078,
+      "step": 13949
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001470602211619621,
+      "loss": 3.2207,
+      "step": 13950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014693301774780244,
+      "loss": 4.1353,
+      "step": 13951
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014680571686494486,
+      "loss": 2.3916,
+      "step": 13952
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014667831877776365,
+      "loss": 0.8317,
+      "step": 13953
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014655082375083362,
+      "loss": 1.0401,
+      "step": 13954
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014642323204893222,
+      "loss": 1.4991,
+      "step": 13955
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014629554393703637,
+      "loss": 2.6201,
+      "step": 13956
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014616775968032453,
+      "loss": 0.2981,
+      "step": 13957
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001460398795441735,
+      "loss": 0.909,
+      "step": 13958
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014591190379416048,
+      "loss": 0.2597,
+      "step": 13959
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014578383269606005,
+      "loss": 3.2485,
+      "step": 13960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014565566651584592,
+      "loss": 1.8708,
+      "step": 13961
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014552740551968812,
+      "loss": 1.8347,
+      "step": 13962
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014539904997395485,
+      "loss": 1.7629,
+      "step": 13963
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001452706001452093,
+      "loss": 1.7218,
+      "step": 13964
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014514205630021185,
+      "loss": 0.6468,
+      "step": 13965
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014501341870591677,
+      "loss": 3.4476,
+      "step": 13966
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001448846876294744,
+      "loss": 1.9404,
+      "step": 13967
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001447558633382278,
+      "loss": 1.6677,
+      "step": 13968
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014462694609971506,
+      "loss": 0.5971,
+      "step": 13969
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014449793618166588,
+      "loss": 2.2933,
+      "step": 13970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014436883385200377,
+      "loss": 1.7471,
+      "step": 13971
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014423963937884287,
+      "loss": 2.2239,
+      "step": 13972
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014411035303049,
+      "loss": 2.7281,
+      "step": 13973
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001439809750754414,
+      "loss": 3.8006,
+      "step": 13974
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000143851505782385,
+      "loss": 0.2229,
+      "step": 13975
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001437219454201973,
+      "loss": 1.8628,
+      "step": 13976
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014359229425794397,
+      "loss": 1.8075,
+      "step": 13977
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014346255256488023,
+      "loss": 2.2383,
+      "step": 13978
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014333272061044808,
+      "loss": 2.0975,
+      "step": 13979
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001432027986642782,
+      "loss": 1.999,
+      "step": 13980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014307278699618684,
+      "loss": 1.1922,
+      "step": 13981
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001429426858761779,
+      "loss": 0.3523,
+      "step": 13982
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001428124955744398,
+      "loss": 0.6876,
+      "step": 13983
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014268221636134744,
+      "loss": 0.2356,
+      "step": 13984
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014255184850745902,
+      "loss": 2.64,
+      "step": 13985
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001424213922835182,
+      "loss": 2.0181,
+      "step": 13986
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014229084796045082,
+      "loss": 1.2083,
+      "step": 13987
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000142160215809367,
+      "loss": 1.4423,
+      "step": 13988
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014202949610155792,
+      "loss": 1.4172,
+      "step": 13989
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014189868910849794,
+      "loss": 1.4675,
+      "step": 13990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014176779510184138,
+      "loss": 4.6503,
+      "step": 13991
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014163681435342456,
+      "loss": 2.4834,
+      "step": 13992
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014150574713526264,
+      "loss": 3.9646,
+      "step": 13993
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014137459371955165,
+      "loss": 0.3741,
+      "step": 13994
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014124335437866537,
+      "loss": 0.2647,
+      "step": 13995
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014111202938515727,
+      "loss": 0.9454,
+      "step": 13996
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014098061901175743,
+      "loss": 3.5215,
+      "step": 13997
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014084912353137455,
+      "loss": 0.7864,
+      "step": 13998
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014071754321709276,
+      "loss": 1.9584,
+      "step": 13999
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001405858783421737,
+      "loss": 1.9162,
+      "step": 14000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014045412918005327,
+      "loss": 1.6397,
+      "step": 14001
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014032229600434342,
+      "loss": 1.5593,
+      "step": 14002
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001401903790888306,
+      "loss": 0.6657,
+      "step": 14003
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00014005837870747448,
+      "loss": 2.0446,
+      "step": 14004
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001399262951344081,
+      "loss": 2.6065,
+      "step": 14005
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001397941286439382,
+      "loss": 3.2861,
+      "step": 14006
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013966187951054278,
+      "loss": 1.9664,
+      "step": 14007
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013952954800887146,
+      "loss": 1.0301,
+      "step": 14008
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001393971344137459,
+      "loss": 0.4882,
+      "step": 14009
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013926463900015696,
+      "loss": 0.5063,
+      "step": 14010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013913206204326666,
+      "loss": 2.2724,
+      "step": 14011
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013899940381840515,
+      "loss": 2.1884,
+      "step": 14012
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013886666460107258,
+      "loss": 3.597,
+      "step": 14013
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013873384466693605,
+      "loss": 1.1642,
+      "step": 14014
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013860094429183156,
+      "loss": 2.4835,
+      "step": 14015
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001384679637517609,
+      "loss": 1.7833,
+      "step": 14016
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001383349033228936,
+      "loss": 1.7333,
+      "step": 14017
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013820176328156384,
+      "loss": 1.4909,
+      "step": 14018
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013806854390427243,
+      "loss": 1.7768,
+      "step": 14019
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001379352454676836,
+      "loss": 3.7316,
+      "step": 14020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001378018682486271,
+      "loss": 3.5309,
+      "step": 14021
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013766841252409504,
+      "loss": 2.0114,
+      "step": 14022
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013753487857124375,
+      "loss": 1.8125,
+      "step": 14023
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013740126666739085,
+      "loss": 0.4422,
+      "step": 14024
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013726757709001705,
+      "loss": 0.4885,
+      "step": 14025
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013713381011676312,
+      "loss": 0.3687,
+      "step": 14026
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013699996602543183,
+      "loss": 2.5919,
+      "step": 14027
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013686604509398486,
+      "loss": 2.5126,
+      "step": 14028
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013673204760054469,
+      "loss": 3.7917,
+      "step": 14029
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013659797382339158,
+      "loss": 3.3,
+      "step": 14030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013646382404096543,
+      "loss": 3.5467,
+      "step": 14031
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013632959853186277,
+      "loss": 0.6252,
+      "step": 14032
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013619529757483866,
+      "loss": 0.3899,
+      "step": 14033
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001360609214488035,
+      "loss": 2.3281,
+      "step": 14034
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013592647043282524,
+      "loss": 1.9729,
+      "step": 14035
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013579194480612582,
+      "loss": 2.4678,
+      "step": 14036
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001356573448480837,
+      "loss": 2.8283,
+      "step": 14037
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013552267083823054,
+      "loss": 0.5897,
+      "step": 14038
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013538792305625182,
+      "loss": 1.7557,
+      "step": 14039
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00013525310178198733,
+      "loss": 1.1705,
+      "step": 14040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013511820729542802,
+      "loss": 0.8883,
+      "step": 14041
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013498323987671827,
+      "loss": 0.3869,
+      "step": 14042
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013484819980615257,
+      "loss": 0.6169,
+      "step": 14043
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001347130873641777,
+      "loss": 1.1988,
+      "step": 14044
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001345779028313893,
+      "loss": 1.5703,
+      "step": 14045
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013444264648853417,
+      "loss": 0.7755,
+      "step": 14046
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013430731861650678,
+      "loss": 2.2156,
+      "step": 14047
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013417191949635158,
+      "loss": 3.0743,
+      "step": 14048
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013403644940925962,
+      "loss": 2.2954,
+      "step": 14049
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013390090863657066,
+      "loss": 0.3754,
+      "step": 14050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013376529745976994,
+      "loss": 0.8011,
+      "step": 14051
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013362961616049021,
+      "loss": 1.9857,
+      "step": 14052
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013349386502050858,
+      "loss": 1.8603,
+      "step": 14053
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013335804432174848,
+      "loss": 0.9098,
+      "step": 14054
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013322215434627646,
+      "loss": 0.2771,
+      "step": 14055
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001330861953763043,
+      "loss": 0.2971,
+      "step": 14056
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013295016769418575,
+      "loss": 0.1778,
+      "step": 14057
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013281407158241857,
+      "loss": 2.0061,
+      "step": 14058
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013267790732364126,
+      "loss": 3.0987,
+      "step": 14059
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013254167520063522,
+      "loss": 3.7486,
+      "step": 14060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001324053754963215,
+      "loss": 3.7287,
+      "step": 14061
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013226900849376273,
+      "loss": 2.3473,
+      "step": 14062
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013213257447616006,
+      "loss": 1.2802,
+      "step": 14063
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013199607372685508,
+      "loss": 1.7516,
+      "step": 14064
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013185950652932666,
+      "loss": 2.3397,
+      "step": 14065
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013172287316719306,
+      "loss": 0.8775,
+      "step": 14066
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013158617392420847,
+      "loss": 2.1825,
+      "step": 14067
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013144940908426539,
+      "loss": 0.6469,
+      "step": 14068
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001313125789313911,
+      "loss": 3.5854,
+      "step": 14069
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013117568374974996,
+      "loss": 2.7681,
+      "step": 14070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013103872382364025,
+      "loss": 1.3432,
+      "step": 14071
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013090169943749482,
+      "loss": 2.7256,
+      "step": 14072
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013076461087588133,
+      "loss": 3.3028,
+      "step": 14073
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001306274584234994,
+      "loss": 2.2091,
+      "step": 14074
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013049024236518267,
+      "loss": 0.7561,
+      "step": 14075
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001303529629858955,
+      "loss": 0.724,
+      "step": 14076
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013021562057073522,
+      "loss": 2.3128,
+      "step": 14077
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00013007821540492854,
+      "loss": 3.5985,
+      "step": 14078
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000129940747773834,
+      "loss": 2.8515,
+      "step": 14079
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001298032179629384,
+      "loss": 1.3864,
+      "step": 14080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001296656262578591,
+      "loss": 2.8534,
+      "step": 14081
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001295279729443406,
+      "loss": 1.4476,
+      "step": 14082
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012939025830825676,
+      "loss": 1.623,
+      "step": 14083
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012925248263560732,
+      "loss": 0.9863,
+      "step": 14084
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001291146462125203,
+      "loss": 1.0378,
+      "step": 14085
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012897674932524833,
+      "loss": 1.3365,
+      "step": 14086
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012883879226017114,
+      "loss": 0.2585,
+      "step": 14087
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012870077530379204,
+      "loss": 0.9696,
+      "step": 14088
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012856269874273997,
+      "loss": 0.6929,
+      "step": 14089
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012842456286376645,
+      "loss": 1.2797,
+      "step": 14090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012828636795374744,
+      "loss": 1.3715,
+      "step": 14091
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012814811429968017,
+      "loss": 2.9015,
+      "step": 14092
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012800980218868524,
+      "loss": 1.5868,
+      "step": 14093
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012787143190800323,
+      "loss": 0.8682,
+      "step": 14094
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012773300374499698,
+      "loss": 1.1487,
+      "step": 14095
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001275945179871481,
+      "loss": 2.9293,
+      "step": 14096
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012745597492205924,
+      "loss": 0.4243,
+      "step": 14097
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012731737483745058,
+      "loss": 5.005,
+      "step": 14098
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012717871802116226,
+      "loss": 2.6176,
+      "step": 14099
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012704000476115073,
+      "loss": 0.9462,
+      "step": 14100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012690123534549104,
+      "loss": 0.7587,
+      "step": 14101
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012676241006237389,
+      "loss": 3.7907,
+      "step": 14102
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012662352920010602,
+      "loss": 3.288,
+      "step": 14103
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001264845930471105,
+      "loss": 1.9917,
+      "step": 14104
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000126345601891924,
+      "loss": 1.7123,
+      "step": 14105
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012620655602319872,
+      "loss": 2.5469,
+      "step": 14106
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012606745572969912,
+      "loss": 1.7369,
+      "step": 14107
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012592830130030402,
+      "loss": 2.3019,
+      "step": 14108
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012578909302400336,
+      "loss": 2.7294,
+      "step": 14109
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001256498311899003,
+      "loss": 3.1748,
+      "step": 14110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012551051608720773,
+      "loss": 2.2948,
+      "step": 14111
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012537114800525068,
+      "loss": 1.4057,
+      "step": 14112
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001252317272334627,
+      "loss": 0.5514,
+      "step": 14113
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012509225406138826,
+      "loss": 0.6756,
+      "step": 14114
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012495272877867926,
+      "loss": 0.903,
+      "step": 14115
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012481315167509708,
+      "loss": 0.403,
+      "step": 14116
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012467352304050948,
+      "loss": 2.8386,
+      "step": 14117
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012453384316489253,
+      "loss": 1.1014,
+      "step": 14118
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012439411233832742,
+      "loss": 3.3355,
+      "step": 14119
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001242543308510024,
+      "loss": 1.7629,
+      "step": 14120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001241144989932097,
+      "loss": 3.2544,
+      "step": 14121
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001239746170553474,
+      "loss": 0.1713,
+      "step": 14122
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012383468532791635,
+      "loss": 1.5525,
+      "step": 14123
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012369470410152208,
+      "loss": 2.5397,
+      "step": 14124
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012355467366687156,
+      "loss": 1.5986,
+      "step": 14125
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012341459431477528,
+      "loss": 0.2969,
+      "step": 14126
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012327446633614406,
+      "loss": 2.1435,
+      "step": 14127
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000123134290021991,
+      "loss": 3.5969,
+      "step": 14128
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012299406566342825,
+      "loss": 2.058,
+      "step": 14129
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000122853793551669,
+      "loss": 1.1896,
+      "step": 14130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012271347397802443,
+      "loss": 3.095,
+      "step": 14131
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001225731072339055,
+      "loss": 0.9456,
+      "step": 14132
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012243269361082018,
+      "loss": 1.2194,
+      "step": 14133
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012229223340037374,
+      "loss": 1.5814,
+      "step": 14134
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012215172689426925,
+      "loss": 0.4058,
+      "step": 14135
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012201117438430463,
+      "loss": 0.6813,
+      "step": 14136
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001218705761623746,
+      "loss": 1.2382,
+      "step": 14137
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012172993252046751,
+      "loss": 1.3302,
+      "step": 14138
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001215892437506674,
+      "loss": 3.6566,
+      "step": 14139
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012144851014515059,
+      "loss": 0.9058,
+      "step": 14140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012130773199618798,
+      "loss": 1.6177,
+      "step": 14141
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012116690959614157,
+      "loss": 2.1045,
+      "step": 14142
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001210260432374666,
+      "loss": 3.4236,
+      "step": 14143
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012088513321270824,
+      "loss": 1.1012,
+      "step": 14144
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012074417981450379,
+      "loss": 5.6198,
+      "step": 14145
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012060318333557914,
+      "loss": 2.6633,
+      "step": 14146
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001204621440687511,
+      "loss": 0.2023,
+      "step": 14147
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012032106230692393,
+      "loss": 0.8689,
+      "step": 14148
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012017993834309153,
+      "loss": 0.2317,
+      "step": 14149
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00012003877247033408,
+      "loss": 5.2367,
+      "step": 14150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011989756498182014,
+      "loss": 0.9705,
+      "step": 14151
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011975631617080336,
+      "loss": 0.345,
+      "step": 14152
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011961502633062453,
+      "loss": 4.0533,
+      "step": 14153
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011947369575470832,
+      "loss": 2.1951,
+      "step": 14154
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011933232473656537,
+      "loss": 1.5556,
+      "step": 14155
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011919091356978889,
+      "loss": 2.2072,
+      "step": 14156
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011904946254805691,
+      "loss": 2.6805,
+      "step": 14157
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011890797196512878,
+      "loss": 1.0774,
+      "step": 14158
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011876644211484744,
+      "loss": 3.7158,
+      "step": 14159
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011862487329113599,
+      "loss": 0.3919,
+      "step": 14160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011848326578799984,
+      "loss": 2.3023,
+      "step": 14161
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011834161989952335,
+      "loss": 2.5769,
+      "step": 14162
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011819993591987197,
+      "loss": 0.4964,
+      "step": 14163
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011805821414328926,
+      "loss": 3.6649,
+      "step": 14164
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001179164548640972,
+      "loss": 0.9434,
+      "step": 14165
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011777465837669675,
+      "loss": 3.3537,
+      "step": 14166
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011763282497556474,
+      "loss": 1.7263,
+      "step": 14167
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011749095495525608,
+      "loss": 4.1021,
+      "step": 14168
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011734904861040028,
+      "loss": 1.3388,
+      "step": 14169
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011720710623570372,
+      "loss": 3.7223,
+      "step": 14170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011706512812594619,
+      "loss": 1.5394,
+      "step": 14171
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011692311457598313,
+      "loss": 2.4147,
+      "step": 14172
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011678106588074214,
+      "loss": 1.0229,
+      "step": 14173
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011663898233522518,
+      "loss": 2.1366,
+      "step": 14174
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001164968642345053,
+      "loss": 1.5242,
+      "step": 14175
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011635471187372859,
+      "loss": 2.6987,
+      "step": 14176
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011621252554811095,
+      "loss": 3.7478,
+      "step": 14177
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011607030555294023,
+      "loss": 1.1242,
+      "step": 14178
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011592805218357275,
+      "loss": 0.3825,
+      "step": 14179
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011578576573543558,
+      "loss": 0.4257,
+      "step": 14180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011564344650402308,
+      "loss": 0.7732,
+      "step": 14181
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001155010947848991,
+      "loss": 1.3087,
+      "step": 14182
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011535871087369354,
+      "loss": 3.3636,
+      "step": 14183
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011521629506610457,
+      "loss": 1.3981,
+      "step": 14184
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011507384765789516,
+      "loss": 1.4651,
+      "step": 14185
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001149313689448954,
+      "loss": 0.6197,
+      "step": 14186
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011478885922299892,
+      "loss": 0.4537,
+      "step": 14187
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011464631878816514,
+      "loss": 1.1209,
+      "step": 14188
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001145037479364159,
+      "loss": 1.5985,
+      "step": 14189
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011436114696383762,
+      "loss": 3.2497,
+      "step": 14190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011421851616657782,
+      "loss": 0.2009,
+      "step": 14191
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011407585584084739,
+      "loss": 0.1298,
+      "step": 14192
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011393316628291715,
+      "loss": 1.0233,
+      "step": 14193
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011379044778912003,
+      "loss": 3.9681,
+      "step": 14194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011364770065584762,
+      "loss": 2.288,
+      "step": 14195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011350492517955245,
+      "loss": 1.4645,
+      "step": 14196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011336212165674478,
+      "loss": 1.2484,
+      "step": 14197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011321929038399317,
+      "loss": 2.5197,
+      "step": 14198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011307643165792486,
+      "loss": 1.4442,
+      "step": 14199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011293354577522269,
+      "loss": 2.6944,
+      "step": 14200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011279063303262729,
+      "loss": 1.0252,
+      "step": 14201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011264769372693371,
+      "loss": 1.9443,
+      "step": 14202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011250472815499349,
+      "loss": 0.4737,
+      "step": 14203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011236173661371141,
+      "loss": 0.6664,
+      "step": 14204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011221871940004751,
+      "loss": 2.359,
+      "step": 14205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011207567681101381,
+      "loss": 1.6218,
+      "step": 14206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011193260914367633,
+      "loss": 1.8934,
+      "step": 14207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001117895166951519,
+      "loss": 4.8492,
+      "step": 14208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011164639976261011,
+      "loss": 3.1851,
+      "step": 14209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011150325864327004,
+      "loss": 1.214,
+      "step": 14210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001113600936344024,
+      "loss": 1.9326,
+      "step": 14211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011121690503332611,
+      "loss": 2.8901,
+      "step": 14212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011107369313741049,
+      "loss": 0.4928,
+      "step": 14213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011093045824407182,
+      "loss": 0.8807,
+      "step": 14214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011078720065077557,
+      "loss": 0.9013,
+      "step": 14215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001106439206550329,
+      "loss": 0.7945,
+      "step": 14216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011050061855440297,
+      "loss": 1.8096,
+      "step": 14217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011035729464648939,
+      "loss": 0.4446,
+      "step": 14218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011021394922894246,
+      "loss": 2.4244,
+      "step": 14219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00011007058259945577,
+      "loss": 2.6205,
+      "step": 14220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00010992719505576838,
+      "loss": 0.5144,
+      "step": 14221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00010978378689566136,
+      "loss": 1.0978,
+      "step": 14222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00010964035841695997,
+      "loss": 2.8674,
+      "step": 14223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001094969099175303,
+      "loss": 2.3807,
+      "step": 14224
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001093534416952814,
+      "loss": 1.7142,
+      "step": 14225
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010920995404816193,
+      "loss": 1.7216,
+      "step": 14226
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010906644727416222,
+      "loss": 6.3797,
+      "step": 14227
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001089229216713113,
+      "loss": 4.4313,
+      "step": 14228
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010877937753767735,
+      "loss": 3.0292,
+      "step": 14229
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010863581517136798,
+      "loss": 2.5814,
+      "step": 14230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010849223487052738,
+      "loss": 3.4361,
+      "step": 14231
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010834863693333828,
+      "loss": 1.7527,
+      "step": 14232
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010820502165801873,
+      "loss": 1.3058,
+      "step": 14233
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010806138934282413,
+      "loss": 2.7005,
+      "step": 14234
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001079177402860439,
+      "loss": 1.0197,
+      "step": 14235
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010777407478600354,
+      "loss": 0.819,
+      "step": 14236
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010763039314106146,
+      "loss": 1.1842,
+      "step": 14237
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010748669564961087,
+      "loss": 2.0114,
+      "step": 14238
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010734298261007653,
+      "loss": 1.34,
+      "step": 14239
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010719925432091691,
+      "loss": 3.1438,
+      "step": 14240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010705551108062067,
+      "loss": 3.2396,
+      "step": 14241
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010691175318770903,
+      "loss": 2.0664,
+      "step": 14242
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010676798094073217,
+      "loss": 0.7709,
+      "step": 14243
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010662419463827151,
+      "loss": 0.1311,
+      "step": 14244
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010648039457893624,
+      "loss": 1.6955,
+      "step": 14245
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010633658106136552,
+      "loss": 2.7692,
+      "step": 14246
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010619275438422507,
+      "loss": 1.0598,
+      "step": 14247
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010604891484620934,
+      "loss": 4.713,
+      "step": 14248
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010590506274603808,
+      "loss": 0.4391,
+      "step": 14249
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010576119838245855,
+      "loss": 2.8231,
+      "step": 14250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010561732205424204,
+      "loss": 1.2353,
+      "step": 14251
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010547343406018612,
+      "loss": 0.4383,
+      "step": 14252
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001053295346991112,
+      "loss": 2.2912,
+      "step": 14253
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010518562426986263,
+      "loss": 0.985,
+      "step": 14254
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001050417030713074,
+      "loss": 0.5912,
+      "step": 14255
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001048977714023363,
+      "loss": 1.7843,
+      "step": 14256
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010475382956186034,
+      "loss": 2.1335,
+      "step": 14257
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010460987784881321,
+      "loss": 1.3083,
+      "step": 14258
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010446591656214795,
+      "loss": 0.915,
+      "step": 14259
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010432194600083747,
+      "loss": 1.9135,
+      "step": 14260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010417796646387508,
+      "loss": 1.4385,
+      "step": 14261
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010403397825027126,
+      "loss": 2.9479,
+      "step": 14262
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010388998165905592,
+      "loss": 1.5127,
+      "step": 14263
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010374597698927502,
+      "loss": 1.7783,
+      "step": 14264
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010360196453999258,
+      "loss": 0.5296,
+      "step": 14265
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001034579446102875,
+      "loss": 1.6231,
+      "step": 14266
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010331391749925558,
+      "loss": 3.5013,
+      "step": 14267
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001031698835060061,
+      "loss": 2.4414,
+      "step": 14268
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010302584292966404,
+      "loss": 1.1688,
+      "step": 14269
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010288179606936667,
+      "loss": 1.1114,
+      "step": 14270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010273774322426573,
+      "loss": 5.2891,
+      "step": 14271
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010259368469352391,
+      "loss": 0.3281,
+      "step": 14272
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010244962077631721,
+      "loss": 2.6404,
+      "step": 14273
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010230555177183137,
+      "loss": 2.3527,
+      "step": 14274
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010216147797926405,
+      "loss": 0.5621,
+      "step": 14275
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010201739969782151,
+      "loss": 2.0586,
+      "step": 14276
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010187331722672075,
+      "loss": 1.546,
+      "step": 14277
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010172923086518602,
+      "loss": 4.4466,
+      "step": 14278
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010158514091245106,
+      "loss": 2.2932,
+      "step": 14279
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010144104766775568,
+      "loss": 2.7417,
+      "step": 14280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010129695143034795,
+      "loss": 1.4719,
+      "step": 14281
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010115285249948067,
+      "loss": 2.773,
+      "step": 14282
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010100875117441375,
+      "loss": 1.7203,
+      "step": 14283
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010086464775441058,
+      "loss": 2.2155,
+      "step": 14284
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010072054253874037,
+      "loss": 0.5576,
+      "step": 14285
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010057643582667458,
+      "loss": 2.533,
+      "step": 14286
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010043232791748927,
+      "loss": 4.4877,
+      "step": 14287
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010028821911046149,
+      "loss": 2.1008,
+      "step": 14288
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00010014410970487165,
+      "loss": 0.3902,
+      "step": 14289
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.999999999999991e-05,
+      "loss": 1.712,
+      "step": 14290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.985589029512853e-05,
+      "loss": 1.3352,
+      "step": 14291
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.971178088953868e-05,
+      "loss": 1.96,
+      "step": 14292
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.95676720825109e-05,
+      "loss": 1.3165,
+      "step": 14293
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.942356417332558e-05,
+      "loss": 3.0902,
+      "step": 14294
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.92794574612598e-05,
+      "loss": 2.6218,
+      "step": 14295
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.913535224558958e-05,
+      "loss": 4.4748,
+      "step": 14296
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.899124882558643e-05,
+      "loss": 3.1728,
+      "step": 14297
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.88471475005195e-05,
+      "loss": 1.3405,
+      "step": 14298
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.870304856965224e-05,
+      "loss": 4.96,
+      "step": 14299
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.855895233224449e-05,
+      "loss": 3.1066,
+      "step": 14300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.84148590875491e-05,
+      "loss": 1.3644,
+      "step": 14301
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.827076913481414e-05,
+      "loss": 1.8893,
+      "step": 14302
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.81266827732794e-05,
+      "loss": 1.3902,
+      "step": 14303
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.798260030217865e-05,
+      "loss": 3.2121,
+      "step": 14304
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.783852202073612e-05,
+      "loss": 0.1677,
+      "step": 14305
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.76944482281688e-05,
+      "loss": 3.7779,
+      "step": 14306
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.755037922368296e-05,
+      "loss": 1.9625,
+      "step": 14307
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.740631530647626e-05,
+      "loss": 4.4477,
+      "step": 14308
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.726225677573446e-05,
+      "loss": 0.8855,
+      "step": 14309
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.71182039306335e-05,
+      "loss": 4.041,
+      "step": 14310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.697415707033612e-05,
+      "loss": 1.1072,
+      "step": 14311
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.683011649399408e-05,
+      "loss": 0.3338,
+      "step": 14312
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.668608250074457e-05,
+      "loss": 2.5699,
+      "step": 14313
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.654205538971264e-05,
+      "loss": 0.154,
+      "step": 14314
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.639803546000759e-05,
+      "loss": 1.1444,
+      "step": 14315
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.625402301072516e-05,
+      "loss": 2.8821,
+      "step": 14316
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.611001834094424e-05,
+      "loss": 0.6914,
+      "step": 14317
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.596602174972891e-05,
+      "loss": 4.1934,
+      "step": 14318
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.58220335361251e-05,
+      "loss": 1.0783,
+      "step": 14319
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.56780539991627e-05,
+      "loss": 2.5222,
+      "step": 14320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.553408343785223e-05,
+      "loss": 2.0062,
+      "step": 14321
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.539012215118696e-05,
+      "loss": 1.4667,
+      "step": 14322
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.524617043813981e-05,
+      "loss": 1.5055,
+      "step": 14323
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.510222859766388e-05,
+      "loss": 0.7759,
+      "step": 14324
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.495829692869276e-05,
+      "loss": 1.3181,
+      "step": 14325
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.481437573013755e-05,
+      "loss": 4.1683,
+      "step": 14326
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.467046530088898e-05,
+      "loss": 2.8037,
+      "step": 14327
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.452656593981404e-05,
+      "loss": 0.675,
+      "step": 14328
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.438267794575813e-05,
+      "loss": 0.7721,
+      "step": 14329
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.423880161754162e-05,
+      "loss": 0.4375,
+      "step": 14330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.409493725396207e-05,
+      "loss": 2.2925,
+      "step": 14331
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.395108515379082e-05,
+      "loss": 0.2678,
+      "step": 14332
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.380724561577509e-05,
+      "loss": 3.6669,
+      "step": 14333
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.366341893863465e-05,
+      "loss": 0.8978,
+      "step": 14334
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.351960542106393e-05,
+      "loss": 1.6931,
+      "step": 14335
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.337580536172865e-05,
+      "loss": 0.8653,
+      "step": 14336
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.323201905926799e-05,
+      "loss": 1.6629,
+      "step": 14337
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.308824681229113e-05,
+      "loss": 1.188,
+      "step": 14338
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.294448891937949e-05,
+      "loss": 1.5737,
+      "step": 14339
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.280074567908327e-05,
+      "loss": 2.581,
+      "step": 14340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.265701738992363e-05,
+      "loss": 2.2971,
+      "step": 14341
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.251330435038928e-05,
+      "loss": 1.7333,
+      "step": 14342
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.23696068589387e-05,
+      "loss": 1.4921,
+      "step": 14343
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.222592521399661e-05,
+      "loss": 1.219,
+      "step": 14344
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.208225971395629e-05,
+      "loss": 0.3273,
+      "step": 14345
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.193861065717604e-05,
+      "loss": 2.7715,
+      "step": 14346
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.179497834198144e-05,
+      "loss": 3.2464,
+      "step": 14347
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.165136306666189e-05,
+      "loss": 2.4175,
+      "step": 14348
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.15077651294728e-05,
+      "loss": 0.7232,
+      "step": 14349
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.136418482863218e-05,
+      "loss": 2.0936,
+      "step": 14350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.122062246232283e-05,
+      "loss": 1.0374,
+      "step": 14351
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.107707832868885e-05,
+      "loss": 2.2823,
+      "step": 14352
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.093355272583795e-05,
+      "loss": 2.3631,
+      "step": 14353
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.079004595183824e-05,
+      "loss": 0.975,
+      "step": 14354
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.064655830471875e-05,
+      "loss": 0.5679,
+      "step": 14355
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.050309008246988e-05,
+      "loss": 3.4414,
+      "step": 14356
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.03596415830402e-05,
+      "loss": 3.8057,
+      "step": 14357
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.021621310433881e-05,
+      "loss": 0.4808,
+      "step": 14358
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.007280494423177e-05,
+      "loss": 2.2549,
+      "step": 14359
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.992941740054438e-05,
+      "loss": 1.3034,
+      "step": 14360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.978605077105771e-05,
+      "loss": 0.9589,
+      "step": 14361
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.964270535351078e-05,
+      "loss": 0.1476,
+      "step": 14362
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.94993814455972e-05,
+      "loss": 2.5145,
+      "step": 14363
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.935607934496727e-05,
+      "loss": 1.8658,
+      "step": 14364
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.92127993492246e-05,
+      "loss": 0.3878,
+      "step": 14365
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.906954175592835e-05,
+      "loss": 1.3332,
+      "step": 14366
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.892630686258968e-05,
+      "loss": 0.5992,
+      "step": 14367
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.878309496667406e-05,
+      "loss": 0.7383,
+      "step": 14368
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.863990636559777e-05,
+      "loss": 1.4335,
+      "step": 14369
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.849674135673013e-05,
+      "loss": 0.692,
+      "step": 14370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.835360023739005e-05,
+      "loss": 1.9431,
+      "step": 14371
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.821048330484827e-05,
+      "loss": 1.5969,
+      "step": 14372
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.806739085632382e-05,
+      "loss": 2.9014,
+      "step": 14373
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.792432318898636e-05,
+      "loss": 1.9469,
+      "step": 14374
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.778128059995265e-05,
+      "loss": 0.6965,
+      "step": 14375
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.763826338628875e-05,
+      "loss": 0.9514,
+      "step": 14376
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.749527184500667e-05,
+      "loss": 2.5834,
+      "step": 14377
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.735230627306646e-05,
+      "loss": 0.2296,
+      "step": 14378
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.720936696737287e-05,
+      "loss": 2.1363,
+      "step": 14379
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.706645422477748e-05,
+      "loss": 1.2429,
+      "step": 14380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.692356834207531e-05,
+      "loss": 3.0204,
+      "step": 14381
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.6780709616007e-05,
+      "loss": 0.8483,
+      "step": 14382
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.663787834325539e-05,
+      "loss": 2.0879,
+      "step": 14383
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.649507482044773e-05,
+      "loss": 1.1919,
+      "step": 14384
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.635229934415254e-05,
+      "loss": 0.9035,
+      "step": 14385
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.620955221088015e-05,
+      "loss": 1.0911,
+      "step": 14386
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.606683371708301e-05,
+      "loss": 1.4907,
+      "step": 14387
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.592414415915278e-05,
+      "loss": 2.996,
+      "step": 14388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.578148383342237e-05,
+      "loss": 2.5218,
+      "step": 14389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.563885303616256e-05,
+      "loss": 0.4759,
+      "step": 14390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.549625206358426e-05,
+      "loss": 1.0553,
+      "step": 14391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.535368121183502e-05,
+      "loss": 3.3549,
+      "step": 14392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.521114077700126e-05,
+      "loss": 0.9732,
+      "step": 14393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.506863105510476e-05,
+      "loss": 2.5237,
+      "step": 14394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.4926152342105e-05,
+      "loss": 1.8853,
+      "step": 14395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.478370493389562e-05,
+      "loss": 0.5943,
+      "step": 14396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.464128912630663e-05,
+      "loss": 2.6237,
+      "step": 14397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.449890521510107e-05,
+      "loss": 0.6985,
+      "step": 14398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.435655349597708e-05,
+      "loss": 4.4664,
+      "step": 14399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.421423426456458e-05,
+      "loss": 1.3296,
+      "step": 14400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.407194781642742e-05,
+      "loss": 1.462,
+      "step": 14401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.392969444705993e-05,
+      "loss": 1.728,
+      "step": 14402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.37874744518892e-05,
+      "loss": 1.7803,
+      "step": 14403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.364528812627157e-05,
+      "loss": 1.0419,
+      "step": 14404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.350313576549486e-05,
+      "loss": 2.5533,
+      "step": 14405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.336101766477497e-05,
+      "loss": 1.4491,
+      "step": 14406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 8.321893411925803e-05,
+      "loss": 2.3108,
+      "step": 14407
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.307688542401704e-05,
+      "loss": 1.9359,
+      "step": 14408
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.293487187405397e-05,
+      "loss": 0.5892,
+      "step": 14409
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.279289376429645e-05,
+      "loss": 0.2382,
+      "step": 14410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.265095138959989e-05,
+      "loss": 1.1935,
+      "step": 14411
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.250904504474408e-05,
+      "loss": 1.7982,
+      "step": 14412
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.236717502443542e-05,
+      "loss": 1.4778,
+      "step": 14413
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.222534162330342e-05,
+      "loss": 1.2818,
+      "step": 14414
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.208354513590298e-05,
+      "loss": 1.2793,
+      "step": 14415
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.194178585671092e-05,
+      "loss": 3.2107,
+      "step": 14416
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.180006408012819e-05,
+      "loss": 1.1465,
+      "step": 14417
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.165838010047682e-05,
+      "loss": 3.2304,
+      "step": 14418
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.151673421200032e-05,
+      "loss": 1.1472,
+      "step": 14419
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.137512670886417e-05,
+      "loss": 0.2199,
+      "step": 14420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.123355788515272e-05,
+      "loss": 2.1942,
+      "step": 14421
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.10920280348714e-05,
+      "loss": 1.8354,
+      "step": 14422
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.095053745194325e-05,
+      "loss": 2.9989,
+      "step": 14423
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.080908643021127e-05,
+      "loss": 1.7831,
+      "step": 14424
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.06676752634348e-05,
+      "loss": 0.6745,
+      "step": 14425
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.052630424529184e-05,
+      "loss": 1.6689,
+      "step": 14426
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.038497366937563e-05,
+      "loss": 0.9164,
+      "step": 14427
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.02436838291968e-05,
+      "loss": 2.7036,
+      "step": 14428
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 8.010243501818002e-05,
+      "loss": 3.062,
+      "step": 14429
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.996122752966608e-05,
+      "loss": 1.1344,
+      "step": 14430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.982006165690863e-05,
+      "loss": 0.378,
+      "step": 14431
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.967893769307624e-05,
+      "loss": 0.605,
+      "step": 14432
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.953785593124908e-05,
+      "loss": 0.4534,
+      "step": 14433
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.939681666442102e-05,
+      "loss": 2.9929,
+      "step": 14434
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.925582018549638e-05,
+      "loss": 1.3167,
+      "step": 14435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.911486678729191e-05,
+      "loss": 0.4153,
+      "step": 14436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.897395676253357e-05,
+      "loss": 3.1084,
+      "step": 14437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.88330904038586e-05,
+      "loss": 0.5495,
+      "step": 14438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.869226800381217e-05,
+      "loss": 4.1669,
+      "step": 14439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.855148985484957e-05,
+      "loss": 4.414,
+      "step": 14440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.841075624933277e-05,
+      "loss": 0.5436,
+      "step": 14441
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.827006747953264e-05,
+      "loss": 1.5663,
+      "step": 14442
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.812942383762554e-05,
+      "loss": 0.8721,
+      "step": 14443
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.798882561569553e-05,
+      "loss": 1.7122,
+      "step": 14444
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.78482731057309e-05,
+      "loss": 0.2894,
+      "step": 14445
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.770776659962644e-05,
+      "loss": 0.744,
+      "step": 14446
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.756730638918e-05,
+      "loss": 2.257,
+      "step": 14447
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.742689276609467e-05,
+      "loss": 3.9355,
+      "step": 14448
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.728652602197574e-05,
+      "loss": 0.2464,
+      "step": 14449
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.714620644833115e-05,
+      "loss": 4.938,
+      "step": 14450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.700593433657194e-05,
+      "loss": 2.1605,
+      "step": 14451
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.686570997800915e-05,
+      "loss": 0.9267,
+      "step": 14452
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.67255336638561e-05,
+      "loss": 4.4012,
+      "step": 14453
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.658540568522489e-05,
+      "loss": 0.5099,
+      "step": 14454
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.644532633312861e-05,
+      "loss": 1.7992,
+      "step": 14455
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.630529589847808e-05,
+      "loss": 0.5153,
+      "step": 14456
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.61653146720838e-05,
+      "loss": 1.8349,
+      "step": 14457
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.602538294465275e-05,
+      "loss": 3.8253,
+      "step": 14458
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.588550100679048e-05,
+      "loss": 1.3347,
+      "step": 14459
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.574566914899777e-05,
+      "loss": 1.1657,
+      "step": 14460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.560588766167277e-05,
+      "loss": 3.0147,
+      "step": 14461
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.546615683510763e-05,
+      "loss": 1.9853,
+      "step": 14462
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.532647695949069e-05,
+      "loss": 1.2455,
+      "step": 14463
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.518684832490308e-05,
+      "loss": 3.7939,
+      "step": 14464
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.504727122132092e-05,
+      "loss": 0.6853,
+      "step": 14465
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.490774593861187e-05,
+      "loss": 0.8716,
+      "step": 14466
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.476827276653747e-05,
+      "loss": 2.482,
+      "step": 14467
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.46288519947495e-05,
+      "loss": 1.2563,
+      "step": 14468
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.448948391279243e-05,
+      "loss": 0.254,
+      "step": 14469
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.435016881009987e-05,
+      "loss": 1.5875,
+      "step": 14470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.421090697599677e-05,
+      "loss": 3.0426,
+      "step": 14471
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.407169869969615e-05,
+      "loss": 4.5587,
+      "step": 14472
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.393254427030105e-05,
+      "loss": 0.4155,
+      "step": 14473
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.379344397680144e-05,
+      "loss": 2.0828,
+      "step": 14474
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.365439810807615e-05,
+      "loss": 1.0162,
+      "step": 14475
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.351540695288966e-05,
+      "loss": 1.5134,
+      "step": 14476
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.337647079989416e-05,
+      "loss": 0.7477,
+      "step": 14477
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.323758993762627e-05,
+      "loss": 2.813,
+      "step": 14478
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.309876465450913e-05,
+      "loss": 0.2379,
+      "step": 14479
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.295999523884944e-05,
+      "loss": 0.3295,
+      "step": 14480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.28212819788379e-05,
+      "loss": 0.25,
+      "step": 14481
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.268262516254957e-05,
+      "loss": 0.4717,
+      "step": 14482
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.254402507794093e-05,
+      "loss": 4.0526,
+      "step": 14483
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.240548201285206e-05,
+      "loss": 1.469,
+      "step": 14484
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.226699625500318e-05,
+      "loss": 1.1866,
+      "step": 14485
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.212856809199693e-05,
+      "loss": 1.6918,
+      "step": 14486
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.199019781131493e-05,
+      "loss": 3.6164,
+      "step": 14487
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.185188570031998e-05,
+      "loss": 1.1646,
+      "step": 14488
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.171363204625272e-05,
+      "loss": 0.1541,
+      "step": 14489
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.15754371362337e-05,
+      "loss": 0.9442,
+      "step": 14490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.143730125726019e-05,
+      "loss": 0.4712,
+      "step": 14491
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.129922469620813e-05,
+      "loss": 2.6457,
+      "step": 14492
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.116120773982902e-05,
+      "loss": 1.1258,
+      "step": 14493
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.102325067475184e-05,
+      "loss": 0.7349,
+      "step": 14494
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.088535378747989e-05,
+      "loss": 1.0858,
+      "step": 14495
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.074751736439284e-05,
+      "loss": 1.7995,
+      "step": 14496
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.060974169174341e-05,
+      "loss": 1.4376,
+      "step": 14497
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.047202705565955e-05,
+      "loss": 1.6555,
+      "step": 14498
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.033437374214106e-05,
+      "loss": 2.2876,
+      "step": 14499
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.019678203706175e-05,
+      "loss": 1.1714,
+      "step": 14500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 7.005925222616616e-05,
+      "loss": 2.1544,
+      "step": 14501
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.992178459507162e-05,
+      "loss": 0.7841,
+      "step": 14502
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.978437942926495e-05,
+      "loss": 3.299,
+      "step": 14503
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.964703701410464e-05,
+      "loss": 0.232,
+      "step": 14504
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.95097576348175e-05,
+      "loss": 4.3917,
+      "step": 14505
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.937254157650077e-05,
+      "loss": 0.3593,
+      "step": 14506
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.923538912411882e-05,
+      "loss": 3.092,
+      "step": 14507
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.909830056250534e-05,
+      "loss": 2.2604,
+      "step": 14508
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.89612761763599e-05,
+      "loss": 0.4386,
+      "step": 14509
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.882431625025021e-05,
+      "loss": 5.4541,
+      "step": 14510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.868742106860903e-05,
+      "loss": 0.3261,
+      "step": 14511
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.855059091573477e-05,
+      "loss": 1.5164,
+      "step": 14512
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.841382607579168e-05,
+      "loss": 3.7274,
+      "step": 14513
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.827712683280711e-05,
+      "loss": 0.3028,
+      "step": 14514
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.814049347067348e-05,
+      "loss": 2.8605,
+      "step": 14515
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.800392627314509e-05,
+      "loss": 0.1271,
+      "step": 14516
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.786742552384011e-05,
+      "loss": 1.6853,
+      "step": 14517
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.773099150623743e-05,
+      "loss": 2.1057,
+      "step": 14518
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.759462450367866e-05,
+      "loss": 0.362,
+      "step": 14519
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.745832479936492e-05,
+      "loss": 0.2319,
+      "step": 14520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.73220926763589e-05,
+      "loss": 0.8284,
+      "step": 14521
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.71859284175816e-05,
+      "loss": 1.2216,
+      "step": 14522
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.704983230581441e-05,
+      "loss": 1.9418,
+      "step": 14523
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.691380462369585e-05,
+      "loss": 2.7545,
+      "step": 14524
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.677784565372371e-05,
+      "loss": 0.935,
+      "step": 14525
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.664195567825169e-05,
+      "loss": 1.2545,
+      "step": 14526
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.650613497949159e-05,
+      "loss": 1.2851,
+      "step": 14527
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.637038383950994e-05,
+      "loss": 0.7228,
+      "step": 14528
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.623470254023023e-05,
+      "loss": 0.6988,
+      "step": 14529
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.60990913634295e-05,
+      "loss": 1.2465,
+      "step": 14530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.596355059074054e-05,
+      "loss": 2.9453,
+      "step": 14531
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.582808050364858e-05,
+      "loss": 1.9884,
+      "step": 14532
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.569268138349338e-05,
+      "loss": 0.2187,
+      "step": 14533
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.555735351146599e-05,
+      "loss": 0.1592,
+      "step": 14534
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.542209716861084e-05,
+      "loss": 1.9844,
+      "step": 14535
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.528691263582246e-05,
+      "loss": 0.1391,
+      "step": 14536
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.515180019384758e-05,
+      "loss": 2.5529,
+      "step": 14537
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.50167601232819e-05,
+      "loss": 1.4427,
+      "step": 14538
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.488179270457215e-05,
+      "loss": 1.6252,
+      "step": 14539
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.474689821801283e-05,
+      "loss": 1.7359,
+      "step": 14540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.461207694374834e-05,
+      "loss": 0.2115,
+      "step": 14541
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.447732916176963e-05,
+      "loss": 2.0964,
+      "step": 14542
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.434265515191647e-05,
+      "loss": 3.7569,
+      "step": 14543
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.420805519387433e-05,
+      "loss": 0.5861,
+      "step": 14544
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.407352956717494e-05,
+      "loss": 0.2469,
+      "step": 14545
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.393907855119663e-05,
+      "loss": 1.5264,
+      "step": 14546
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.38047024251615e-05,
+      "loss": 0.432,
+      "step": 14547
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.367040146813739e-05,
+      "loss": 0.2653,
+      "step": 14548
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.353617595903472e-05,
+      "loss": 0.9345,
+      "step": 14549
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.340202617660858e-05,
+      "loss": 3.5669,
+      "step": 14550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.326795239945546e-05,
+      "loss": 1.3652,
+      "step": 14551
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.31339549060153e-05,
+      "loss": 0.923,
+      "step": 14552
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.30000339745683e-05,
+      "loss": 0.7563,
+      "step": 14553
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.286618988323703e-05,
+      "loss": 2.3109,
+      "step": 14554
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.27324229099831e-05,
+      "loss": 3.4957,
+      "step": 14555
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.25987333326093e-05,
+      "loss": 0.2316,
+      "step": 14556
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.24651214287564e-05,
+      "loss": 0.2598,
+      "step": 14557
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.233158747590513e-05,
+      "loss": 0.6299,
+      "step": 14558
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.219813175137304e-05,
+      "loss": 1.4599,
+      "step": 14559
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.206475453231657e-05,
+      "loss": 3.1226,
+      "step": 14560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.193145609572774e-05,
+      "loss": 0.6905,
+      "step": 14561
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.179823671843632e-05,
+      "loss": 1.1886,
+      "step": 14562
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.166509667710656e-05,
+      "loss": 1.8493,
+      "step": 14563
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.153203624823927e-05,
+      "loss": 1.9012,
+      "step": 14564
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.13990557081686e-05,
+      "loss": 1.071,
+      "step": 14565
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.126615533306411e-05,
+      "loss": 0.5619,
+      "step": 14566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.113333539892758e-05,
+      "loss": 0.7069,
+      "step": 14567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.1000596181595017e-05,
+      "loss": 2.5461,
+      "step": 14568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.086793795673349e-05,
+      "loss": 3.0493,
+      "step": 14569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0735360999843205e-05,
+      "loss": 0.5777,
+      "step": 14570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0602865586254244e-05,
+      "loss": 1.764,
+      "step": 14571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0470451991128684e-05,
+      "loss": 0.9227,
+      "step": 14572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.033812048945737e-05,
+      "loss": 0.7941,
+      "step": 14573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.020587135606195e-05,
+      "loss": 0.9911,
+      "step": 14574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0073704865592054e-05,
+      "loss": 1.8247,
+      "step": 14575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.994162129252566e-05,
+      "loss": 1.818,
+      "step": 14576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.980962091116955e-05,
+      "loss": 2.3782,
+      "step": 14577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.967770399565673e-05,
+      "loss": 1.9074,
+      "step": 14578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.9545870819946893e-05,
+      "loss": 3.7563,
+      "step": 14579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.941412165782646e-05,
+      "loss": 1.006,
+      "step": 14580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.928245678290738e-05,
+      "loss": 0.9853,
+      "step": 14581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.9150876468625604e-05,
+      "loss": 1.9707,
+      "step": 14582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.901938098824272e-05,
+      "loss": 1.6756,
+      "step": 14583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.888797061484288e-05,
+      "loss": 1.9194,
+      "step": 14584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.875664562133478e-05,
+      "loss": 0.3704,
+      "step": 14585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.8625406280448495e-05,
+      "loss": 2.1263,
+      "step": 14586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.849425286473751e-05,
+      "loss": 3.9451,
+      "step": 14587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.836318564657559e-05,
+      "loss": 2.6822,
+      "step": 14588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.823220489815877e-05,
+      "loss": 3.1314,
+      "step": 14589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.810131089150222e-05,
+      "loss": 2.6081,
+      "step": 14590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.7970503898442255e-05,
+      "loss": 1.4281,
+      "step": 14591
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.783978419063317e-05,
+      "loss": 0.8543,
+      "step": 14592
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.770915203954934e-05,
+      "loss": 2.9256,
+      "step": 14593
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.757860771648196e-05,
+      "loss": 0.6894,
+      "step": 14594
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.744815149254114e-05,
+      "loss": 0.1462,
+      "step": 14595
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.731778363865272e-05,
+      "loss": 0.5966,
+      "step": 14596
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.718750442556034e-05,
+      "loss": 2.068,
+      "step": 14597
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.705731412382225e-05,
+      "loss": 1.0884,
+      "step": 14598
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.692721300381333e-05,
+      "loss": 1.5653,
+      "step": 14599
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.6797201335721964e-05,
+      "loss": 2.4928,
+      "step": 14600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.666727938955205e-05,
+      "loss": 0.2378,
+      "step": 14601
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.6537447435119904e-05,
+      "loss": 1.1505,
+      "step": 14602
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.64077057420562e-05,
+      "loss": 1.3032,
+      "step": 14603
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.6278054579802875e-05,
+      "loss": 2.2251,
+      "step": 14604
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.6148494217615154e-05,
+      "loss": 4.545,
+      "step": 14605
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.601902492455873e-05,
+      "loss": 1.2984,
+      "step": 14606
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.588964696951016e-05,
+      "loss": 0.3279,
+      "step": 14607
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.576036062115728e-05,
+      "loss": 3.0696,
+      "step": 14608
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.563116614799639e-05,
+      "loss": 1.1441,
+      "step": 14609
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.550206381833428e-05,
+      "loss": 0.5013,
+      "step": 14610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.537305390028509e-05,
+      "loss": 1.2206,
+      "step": 14611
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.524413666177233e-05,
+      "loss": 0.4807,
+      "step": 14612
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.5115312370525764e-05,
+      "loss": 0.2183,
+      "step": 14613
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.498658129408336e-05,
+      "loss": 1.5122,
+      "step": 14614
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.485794369978832e-05,
+      "loss": 0.2347,
+      "step": 14615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.472939985479085e-05,
+      "loss": 4.215,
+      "step": 14616
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.460095002604532e-05,
+      "loss": 1.9049,
+      "step": 14617
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.447259448031202e-05,
+      "loss": 0.1605,
+      "step": 14618
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.434433348415423e-05,
+      "loss": 0.1264,
+      "step": 14619
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.42161673039401e-05,
+      "loss": 2.9258,
+      "step": 14620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.4088096205839656e-05,
+      "loss": 1.8205,
+      "step": 14621
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.3960120455826654e-05,
+      "loss": 0.298,
+      "step": 14622
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.383224031967562e-05,
+      "loss": 1.0028,
+      "step": 14623
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.370445606296377e-05,
+      "loss": 1.3797,
+      "step": 14624
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.357676795106794e-05,
+      "loss": 2.2537,
+      "step": 14625
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.344917624916654e-05,
+      "loss": 2.5529,
+      "step": 14626
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.3321681222236505e-05,
+      "loss": 2.7154,
+      "step": 14627
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.319428313505528e-05,
+      "loss": 0.1925,
+      "step": 14628
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.306698225219771e-05,
+      "loss": 1.1444,
+      "step": 14629
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.2939778838038024e-05,
+      "loss": 1.0695,
+      "step": 14630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.2812673156746804e-05,
+      "loss": 0.6783,
+      "step": 14631
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.268566547229285e-05,
+      "loss": 2.0627,
+      "step": 14632
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.2558756048440305e-05,
+      "loss": 3.2489,
+      "step": 14633
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.2431945148750395e-05,
+      "loss": 1.6291,
+      "step": 14634
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.230523303657853e-05,
+      "loss": 0.8557,
+      "step": 14635
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.2178619975076215e-05,
+      "loss": 0.288,
+      "step": 14636
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.20521062271883e-05,
+      "loss": 1.9343,
+      "step": 14637
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.1925692055653386e-05,
+      "loss": 0.3852,
+      "step": 14638
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.1799377723004225e-05,
+      "loss": 1.8973,
+      "step": 14639
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.167316349156497e-05,
+      "loss": 1.3253,
+      "step": 14640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.1547049623453126e-05,
+      "loss": 2.5336,
+      "step": 14641
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.142103638057685e-05,
+      "loss": 1.5807,
+      "step": 14642
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.129512402463533e-05,
+      "loss": 1.7708,
+      "step": 14643
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.116931281711883e-05,
+      "loss": 1.4048,
+      "step": 14644
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.1043603019307576e-05,
+      "loss": 0.5808,
+      "step": 14645
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.0917994892270246e-05,
+      "loss": 1.696,
+      "step": 14646
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.0792488696865626e-05,
+      "loss": 1.703,
+      "step": 14647
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.066708469373955e-05,
+      "loss": 0.5161,
+      "step": 14648
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.054178314332689e-05,
+      "loss": 0.1199,
+      "step": 14649
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.041658430584851e-05,
+      "loss": 1.6764,
+      "step": 14650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.029148844131317e-05,
+      "loss": 1.5729,
+      "step": 14651
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.016649580951456e-05,
+      "loss": 1.9563,
+      "step": 14652
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.004160667003324e-05,
+      "loss": 1.2643,
+      "step": 14653
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.991682128223356e-05,
+      "loss": 1.5813,
+      "step": 14654
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.979213990526569e-05,
+      "loss": 0.7409,
+      "step": 14655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.96675627980625e-05,
+      "loss": 0.3025,
+      "step": 14656
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.9543090219341604e-05,
+      "loss": 0.6076,
+      "step": 14657
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.941872242760226e-05,
+      "loss": 2.8415,
+      "step": 14658
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.929445968112739e-05,
+      "loss": 2.5517,
+      "step": 14659
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.917030223798047e-05,
+      "loss": 0.3375,
+      "step": 14660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.904625035600757e-05,
+      "loss": 2.7233,
+      "step": 14661
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.892230429283429e-05,
+      "loss": 1.7508,
+      "step": 14662
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.8798464305867655e-05,
+      "loss": 2.1358,
+      "step": 14663
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.8674730652293234e-05,
+      "loss": 0.2456,
+      "step": 14664
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.855110358907695e-05,
+      "loss": 1.2529,
+      "step": 14665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.842758337296216e-05,
+      "loss": 0.3139,
+      "step": 14666
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.830417026047154e-05,
+      "loss": 1.0958,
+      "step": 14667
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.81808645079041e-05,
+      "loss": 2.5665,
+      "step": 14668
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.805766637133713e-05,
+      "loss": 2.4405,
+      "step": 14669
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.793457610662352e-05,
+      "loss": 0.3859,
+      "step": 14670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.781159396939211e-05,
+      "loss": 3.1419,
+      "step": 14671
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.768872021504811e-05,
+      "loss": 2.6035,
+      "step": 14672
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.7565955098770745e-05,
+      "loss": 1.5532,
+      "step": 14673
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.744329887551361e-05,
+      "loss": 1.0988,
+      "step": 14674
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.7320751800004745e-05,
+      "loss": 1.4989,
+      "step": 14675
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.719831412674557e-05,
+      "loss": 1.4398,
+      "step": 14676
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.707598611000933e-05,
+      "loss": 0.7983,
+      "step": 14677
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.695376800384279e-05,
+      "loss": 0.6786,
+      "step": 14678
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.6831660062063254e-05,
+      "loss": 1.0192,
+      "step": 14679
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.67096625382604e-05,
+      "loss": 2.1955,
+      "step": 14680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.6587775685793456e-05,
+      "loss": 4.1065,
+      "step": 14681
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.646599975779295e-05,
+      "loss": 2.4975,
+      "step": 14682
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.634433500715789e-05,
+      "loss": 0.1198,
+      "step": 14683
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.622278168655757e-05,
+      "loss": 1.7588,
+      "step": 14684
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.61013400484287e-05,
+      "loss": 1.5844,
+      "step": 14685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.598001034497722e-05,
+      "loss": 1.5672,
+      "step": 14686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.5858792828175426e-05,
+      "loss": 2.0574,
+      "step": 14687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.57376877497638e-05,
+      "loss": 0.8117,
+      "step": 14688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.561669536124819e-05,
+      "loss": 1.5013,
+      "step": 14689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.5495815913901516e-05,
+      "loss": 0.2329,
+      "step": 14690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.5375049658761015e-05,
+      "loss": 1.2476,
+      "step": 14691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.525439684663003e-05,
+      "loss": 4.8354,
+      "step": 14692
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.513385772807511e-05,
+      "loss": 1.5662,
+      "step": 14693
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.50134325534279e-05,
+      "loss": 2.7673,
+      "step": 14694
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.489312157278218e-05,
+      "loss": 2.3643,
+      "step": 14695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4772925035995805e-05,
+      "loss": 0.3838,
+      "step": 14696
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4652843192687756e-05,
+      "loss": 1.8573,
+      "step": 14697
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.453287629224e-05,
+      "loss": 3.0806,
+      "step": 14698
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4413024583794625e-05,
+      "loss": 0.1075,
+      "step": 14699
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4293288316255674e-05,
+      "loss": 0.8621,
+      "step": 14700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.417366773828655e-05,
+      "loss": 2.3646,
+      "step": 14701
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.4054163098310406e-05,
+      "loss": 1.2393,
+      "step": 14702
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.393477464451049e-05,
+      "loss": 1.6192,
+      "step": 14703
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.381550262482791e-05,
+      "loss": 0.506,
+      "step": 14704
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3696347286961914e-05,
+      "loss": 0.8729,
+      "step": 14705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.357730887837036e-05,
+      "loss": 0.0771,
+      "step": 14706
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.345838764626707e-05,
+      "loss": 2.1029,
+      "step": 14707
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.333958383762342e-05,
+      "loss": 0.383,
+      "step": 14708
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3220897699166904e-05,
+      "loss": 1.0698,
+      "step": 14709
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.3102329477379775e-05,
+      "loss": 2.0211,
+      "step": 14710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2983879418500564e-05,
+      "loss": 1.7229,
+      "step": 14711
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.286554776852124e-05,
+      "loss": 1.2709,
+      "step": 14712
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.274733477318905e-05,
+      "loss": 0.1635,
+      "step": 14713
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2629240678003624e-05,
+      "loss": 0.3688,
+      "step": 14714
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.251126572821886e-05,
+      "loss": 0.636,
+      "step": 14715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.239341016884004e-05,
+      "loss": 0.9915,
+      "step": 14716
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2275674244625664e-05,
+      "loss": 1.2915,
+      "step": 14717
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.215805820008463e-05,
+      "loss": 0.4158,
+      "step": 14718
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.2040562279478e-05,
+      "loss": 0.7808,
+      "step": 14719
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.192318672681625e-05,
+      "loss": 2.554,
+      "step": 14720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.180593178586103e-05,
+      "loss": 1.8148,
+      "step": 14721
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.168879770012233e-05,
+      "loss": 0.8537,
+      "step": 14722
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.1571784712860316e-05,
+      "loss": 2.7944,
+      "step": 14723
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.145489306708251e-05,
+      "loss": 1.8183,
+      "step": 14724
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.1338123005545605e-05,
+      "loss": 1.0032,
+      "step": 14725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.122147477075261e-05,
+      "loss": 1.5612,
+      "step": 14726
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.110494860495471e-05,
+      "loss": 1.4758,
+      "step": 14727
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.098854475014841e-05,
+      "loss": 0.7852,
+      "step": 14728
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.087226344807736e-05,
+      "loss": 1.7573,
+      "step": 14729
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.0756104940229555e-05,
+      "loss": 0.8931,
+      "step": 14730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.0640069467839125e-05,
+      "loss": 1.0917,
+      "step": 14731
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.05241572718838e-05,
+      "loss": 0.147,
+      "step": 14732
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.0408368593085324e-05,
+      "loss": 1.7503,
+      "step": 14733
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.029270367190977e-05,
+      "loss": 1.5777,
+      "step": 14734
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.017716274856506e-05,
+      "loss": 4.1518,
+      "step": 14735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.006174606300271e-05,
+      "loss": 0.67,
+      "step": 14736
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.994645385491544e-05,
+      "loss": 3.5355,
+      "step": 14737
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9831286363737395e-05,
+      "loss": 1.4758,
+      "step": 14738
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9716243828644304e-05,
+      "loss": 1.0252,
+      "step": 14739
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.960132648855239e-05,
+      "loss": 2.237,
+      "step": 14740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.948653458211703e-05,
+      "loss": 4.3795,
+      "step": 14741
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.937186834773423e-05,
+      "loss": 1.4624,
+      "step": 14742
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.925732802353788e-05,
+      "loss": 2.1116,
+      "step": 14743
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9142913847401505e-05,
+      "loss": 1.1217,
+      "step": 14744
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.902862605693555e-05,
+      "loss": 0.2379,
+      "step": 14745
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8914464889489074e-05,
+      "loss": 2.0677,
+      "step": 14746
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8800430582147073e-05,
+      "loss": 0.0925,
+      "step": 14747
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.86865233717322e-05,
+      "loss": 0.3089,
+      "step": 14748
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.857274349480203e-05,
+      "loss": 3.7194,
+      "step": 14749
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.845909118765083e-05,
+      "loss": 3.1295,
+      "step": 14750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.83455666863068e-05,
+      "loss": 0.5676,
+      "step": 14751
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8232170226533847e-05,
+      "loss": 0.121,
+      "step": 14752
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.811890204382885e-05,
+      "loss": 2.1793,
+      "step": 14753
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.80057623734234e-05,
+      "loss": 3.3292,
+      "step": 14754
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7892751450281085e-05,
+      "loss": 2.0501,
+      "step": 14755
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.777986950909922e-05,
+      "loss": 1.374,
+      "step": 14756
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7667116784306154e-05,
+      "loss": 3.2417,
+      "step": 14757
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7554493510062995e-05,
+      "loss": 0.4588,
+      "step": 14758
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7441999920260906e-05,
+      "loss": 2.0935,
+      "step": 14759
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7329636248522815e-05,
+      "loss": 2.0176,
+      "step": 14760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.721740272820074e-05,
+      "loss": 5.1096,
+      "step": 14761
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.710529959237755e-05,
+      "loss": 1.9436,
+      "step": 14762
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.6993327073864184e-05,
+      "loss": 1.27,
+      "step": 14763
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.6881485405201434e-05,
+      "loss": 1.9374,
+      "step": 14764
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.676977481865753e-05,
+      "loss": 2.0113,
+      "step": 14765
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.665819554622849e-05,
+      "loss": 3.7693,
+      "step": 14766
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.654674781963841e-05,
+      "loss": 0.6678,
+      "step": 14767
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.6435431870337424e-05,
+      "loss": 2.5791,
+      "step": 14768
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.6324247929501934e-05,
+      "loss": 1.035,
+      "step": 14769
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.621319622803505e-05,
+      "loss": 2.1014,
+      "step": 14770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.6102276996564176e-05,
+      "loss": 0.8327,
+      "step": 14771
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.5991490465442404e-05,
+      "loss": 1.6631,
+      "step": 14772
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.588083686474723e-05,
+      "loss": 1.9193,
+      "step": 14773
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.577031642427927e-05,
+      "loss": 1.9664,
+      "step": 14774
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.565992937356371e-05,
+      "loss": 1.9023,
+      "step": 14775
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.554967594184761e-05,
+      "loss": 1.3978,
+      "step": 14776
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.543955635810162e-05,
+      "loss": 0.4317,
+      "step": 14777
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.5329570851017356e-05,
+      "loss": 3.0659,
+      "step": 14778
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.5219719649009054e-05,
+      "loss": 0.7864,
+      "step": 14779
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.5110002980210934e-05,
+      "loss": 1.4254,
+      "step": 14780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.500042107247893e-05,
+      "loss": 0.7207,
+      "step": 14781
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.489097415338802e-05,
+      "loss": 1.6305,
+      "step": 14782
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.478166245023394e-05,
+      "loss": 0.9858,
+      "step": 14783
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.467248619003049e-05,
+      "loss": 1.1915,
+      "step": 14784
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.456344559951131e-05,
+      "loss": 2.9248,
+      "step": 14785
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.44545409051272e-05,
+      "loss": 0.3801,
+      "step": 14786
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.4345772333047774e-05,
+      "loss": 1.3319,
+      "step": 14787
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.423714010915894e-05,
+      "loss": 1.1202,
+      "step": 14788
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.412864445906447e-05,
+      "loss": 0.2254,
+      "step": 14789
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.402028560808346e-05,
+      "loss": 1.8042,
+      "step": 14790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.3912063781251944e-05,
+      "loss": 0.1777,
+      "step": 14791
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.3803979203320365e-05,
+      "loss": 2.3917,
+      "step": 14792
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.3696032098755156e-05,
+      "loss": 3.0851,
+      "step": 14793
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.3588222691736224e-05,
+      "loss": 2.1972,
+      "step": 14794
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.348055120615856e-05,
+      "loss": 0.1153,
+      "step": 14795
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.3373017865629895e-05,
+      "loss": 0.1496,
+      "step": 14796
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.3265622893471124e-05,
+      "loss": 0.9009,
+      "step": 14797
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.315836651271654e-05,
+      "loss": 2.684,
+      "step": 14798
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.305124894611183e-05,
+      "loss": 2.228,
+      "step": 14799
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.294427041611441e-05,
+      "loss": 0.9668,
+      "step": 14800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.283743114489372e-05,
+      "loss": 1.9361,
+      "step": 14801
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.273073135432894e-05,
+      "loss": 0.2986,
+      "step": 14802
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.26241712660104e-05,
+      "loss": 2.5057,
+      "step": 14803
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.251775110123826e-05,
+      "loss": 1.0346,
+      "step": 14804
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.241147108102134e-05,
+      "loss": 0.8525,
+      "step": 14805
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.230533142607846e-05,
+      "loss": 0.8133,
+      "step": 14806
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.219933235683584e-05,
+      "loss": 2.4417,
+      "step": 14807
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.2093474093428855e-05,
+      "loss": 0.451,
+      "step": 14808
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.198775685569936e-05,
+      "loss": 4.076,
+      "step": 14809
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.1882180863197395e-05,
+      "loss": 0.7694,
+      "step": 14810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.1776746335178645e-05,
+      "loss": 1.1971,
+      "step": 14811
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.1671453490606e-05,
+      "loss": 2.1789,
+      "step": 14812
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.156630254814711e-05,
+      "loss": 0.1852,
+      "step": 14813
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.1461293726175946e-05,
+      "loss": 1.0568,
+      "step": 14814
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.135642724277031e-05,
+      "loss": 0.5686,
+      "step": 14815
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.1251703315713424e-05,
+      "loss": 2.8538,
+      "step": 14816
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.1147122162491405e-05,
+      "loss": 0.1214,
+      "step": 14817
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.104268400029493e-05,
+      "loss": 0.3713,
+      "step": 14818
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.093838904601665e-05,
+      "loss": 0.182,
+      "step": 14819
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.083423751625287e-05,
+      "loss": 0.255,
+      "step": 14820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.0730229627300985e-05,
+      "loss": 1.0575,
+      "step": 14821
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.0626365595161113e-05,
+      "loss": 1.1486,
+      "step": 14822
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.0522645635533606e-05,
+      "loss": 1.835,
+      "step": 14823
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.0419069963820613e-05,
+      "loss": 1.984,
+      "step": 14824
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.0315638795123623e-05,
+      "loss": 2.1343,
+      "step": 14825
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.0212352344245043e-05,
+      "loss": 0.844,
+      "step": 14826
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.0109210825685972e-05,
+      "loss": 2.2923,
+      "step": 14827
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.000621445364653e-05,
+      "loss": 1.1676,
+      "step": 14828
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9903363442026144e-05,
+      "loss": 0.6707,
+      "step": 14829
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.980065800442161e-05,
+      "loss": 1.8966,
+      "step": 14830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9698098354127403e-05,
+      "loss": 3.0118,
+      "step": 14831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9595684704135997e-05,
+      "loss": 1.8441,
+      "step": 14832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9493417267135626e-05,
+      "loss": 2.7562,
+      "step": 14833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9391296255511913e-05,
+      "loss": 2.2353,
+      "step": 14834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9289321881345366e-05,
+      "loss": 0.249,
+      "step": 14835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9187494356412726e-05,
+      "loss": 0.7428,
+      "step": 14836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.9085813892185744e-05,
+      "loss": 1.2908,
+      "step": 14837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.8984280699830024e-05,
+      "loss": 0.8637,
+      "step": 14838
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.888289499020632e-05,
+      "loss": 3.8677,
+      "step": 14839
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.878165697386811e-05,
+      "loss": 0.8255,
+      "step": 14840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.8680566861063118e-05,
+      "loss": 0.7262,
+      "step": 14841
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.8579624861730936e-05,
+      "loss": 0.4207,
+      "step": 14842
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.8478831185504563e-05,
+      "loss": 2.2462,
+      "step": 14843
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.8378186041707933e-05,
+      "loss": 1.9837,
+      "step": 14844
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.827768963935754e-05,
+      "loss": 0.9232,
+      "step": 14845
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.8177342187159984e-05,
+      "loss": 4.7348,
+      "step": 14846
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.807714389351349e-05,
+      "loss": 0.1313,
+      "step": 14847
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.797709496650558e-05,
+      "loss": 0.1413,
+      "step": 14848
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.787719561391452e-05,
+      "loss": 3.526,
+      "step": 14849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.777744604320698e-05,
+      "loss": 0.3032,
+      "step": 14850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7677846461539546e-05,
+      "loss": 1.607,
+      "step": 14851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7578397075756325e-05,
+      "loss": 0.2718,
+      "step": 14852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.74790980923905e-05,
+      "loss": 0.1712,
+      "step": 14853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.737994971766191e-05,
+      "loss": 1.8753,
+      "step": 14854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.728095215747861e-05,
+      "loss": 3.0111,
+      "step": 14855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.718210561743444e-05,
+      "loss": 2.2073,
+      "step": 14856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.708341030281063e-05,
+      "loss": 1.3385,
+      "step": 14857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6984866418573572e-05,
+      "loss": 1.2075,
+      "step": 14858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6886474169375208e-05,
+      "loss": 2.0372,
+      "step": 14859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.678823375955327e-05,
+      "loss": 1.0125,
+      "step": 14860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6690145393129228e-05,
+      "loss": 2.7474,
+      "step": 14861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.659220927380972e-05,
+      "loss": 1.5015,
+      "step": 14862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6494425604984497e-05,
+      "loss": 0.9896,
+      "step": 14863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6396794589726704e-05,
+      "loss": 0.8671,
+      "step": 14864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6299316430793186e-05,
+      "loss": 1.0897,
+      "step": 14865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6201991330622377e-05,
+      "loss": 1.1051,
+      "step": 14866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6104819491335796e-05,
+      "loss": 1.3882,
+      "step": 14867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.600780111473574e-05,
+      "loss": 1.0674,
+      "step": 14868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5910936402306496e-05,
+      "loss": 0.6671,
+      "step": 14869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5814225555213234e-05,
+      "loss": 2.2837,
+      "step": 14870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5717668774300886e-05,
+      "loss": 0.1278,
+      "step": 14871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5621266260095333e-05,
+      "loss": 0.3017,
+      "step": 14872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5525018212801178e-05,
+      "loss": 0.1178,
+      "step": 14873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5428924832303158e-05,
+      "loss": 3.5045,
+      "step": 14874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5332986318163855e-05,
+      "loss": 1.5593,
+      "step": 14875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5237202869625178e-05,
+      "loss": 1.8814,
+      "step": 14876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.514157468560606e-05,
+      "loss": 0.8425,
+      "step": 14877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.504610196470394e-05,
+      "loss": 2.6052,
+      "step": 14878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4950784905192447e-05,
+      "loss": 1.9376,
+      "step": 14879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.485562370502287e-05,
+      "loss": 2.8341,
+      "step": 14880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4760618561821867e-05,
+      "loss": 0.1671,
+      "step": 14881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.466576967289299e-05,
+      "loss": 1.7593,
+      "step": 14882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.45710772352143e-05,
+      "loss": 0.9838,
+      "step": 14883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4476541445439903e-05,
+      "loss": 3.3961,
+      "step": 14884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4382162499897644e-05,
+      "loss": 0.6947,
+      "step": 14885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4287940594590587e-05,
+      "loss": 1.8363,
+      "step": 14886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.419387592519471e-05,
+      "loss": 3.2695,
+      "step": 14887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4099968687060383e-05,
+      "loss": 4.6846,
+      "step": 14888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.400621907521011e-05,
+      "loss": 0.189,
+      "step": 14889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.391262728433997e-05,
+      "loss": 1.3366,
+      "step": 14890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3819193508817582e-05,
+      "loss": 0.3028,
+      "step": 14891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3725917942682407e-05,
+      "loss": 1.5866,
+      "step": 14892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3632800779646047e-05,
+      "loss": 0.2797,
+      "step": 14893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3539842213090423e-05,
+      "loss": 2.4622,
+      "step": 14894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3447042436068112e-05,
+      "loss": 0.1981,
+      "step": 14895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3354401641302604e-05,
+      "loss": 3.2832,
+      "step": 14896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.32619200211863e-05,
+      "loss": 2.2722,
+      "step": 14897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3169597767781957e-05,
+      "loss": 2.3643,
+      "step": 14898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.307743507282044e-05,
+      "loss": 2.7304,
+      "step": 14899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.2985432127701946e-05,
+      "loss": 0.3058,
+      "step": 14900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.2893589123494887e-05,
+      "loss": 0.8227,
+      "step": 14901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.2801906250934858e-05,
+      "loss": 3.1261,
+      "step": 14902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.2710383700425807e-05,
+      "loss": 2.9071,
+      "step": 14903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.261902166203782e-05,
+      "loss": 0.5941,
+      "step": 14904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.252782032550852e-05,
+      "loss": 0.2797,
+      "step": 14905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.243677988024091e-05,
+      "loss": 3.2068,
+      "step": 14906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.2345900515304764e-05,
+      "loss": 0.5676,
+      "step": 14907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.225518241943443e-05,
+      "loss": 1.7462,
+      "step": 14908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.2164625781030225e-05,
+      "loss": 0.6702,
+      "step": 14909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.2074230788156258e-05,
+      "loss": 1.5345,
+      "step": 14910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1983997628541852e-05,
+      "loss": 1.711,
+      "step": 14911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.189392648957933e-05,
+      "loss": 1.8756,
+      "step": 14912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.180401755832545e-05,
+      "loss": 1.4448,
+      "step": 14913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.17142710214992e-05,
+      "loss": 0.1146,
+      "step": 14914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1624687065483183e-05,
+      "loss": 0.605,
+      "step": 14915
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1535265876321508e-05,
+      "loss": 0.981,
+      "step": 14916
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1446007639721122e-05,
+      "loss": 0.7588,
+      "step": 14917
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.135691254104968e-05,
+      "loss": 0.7275,
+      "step": 14918
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1267980765336913e-05,
+      "loss": 0.1882,
+      "step": 14919
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1179212497272493e-05,
+      "loss": 0.1625,
+      "step": 14920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1090607921207416e-05,
+      "loss": 0.1079,
+      "step": 14921
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.1002167221152046e-05,
+      "loss": 1.093,
+      "step": 14922
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.091389058077644e-05,
+      "loss": 0.681,
+      "step": 14923
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.082577818341058e-05,
+      "loss": 2.7737,
+      "step": 14924
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.073783021204272e-05,
+      "loss": 0.2293,
+      "step": 14925
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0650046849319614e-05,
+      "loss": 0.135,
+      "step": 14926
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.056242827754684e-05,
+      "loss": 3.2197,
+      "step": 14927
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0474974678686865e-05,
+      "loss": 4.5809,
+      "step": 14928
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0387686234360403e-05,
+      "loss": 1.5596,
+      "step": 14929
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.030056312584434e-05,
+      "loss": 1.9946,
+      "step": 14930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0213605534073043e-05,
+      "loss": 1.2935,
+      "step": 14931
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0126813639636277e-05,
+      "loss": 3.2019,
+      "step": 14932
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.004018762278036e-05,
+      "loss": 0.2662,
+      "step": 14933
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9953727663407106e-05,
+      "loss": 0.1103,
+      "step": 14934
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.986743394107282e-05,
+      "loss": 0.4855,
+      "step": 14935
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.978130663498944e-05,
+      "loss": 0.1363,
+      "step": 14936
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9695345924022436e-05,
+      "loss": 0.0745,
+      "step": 14937
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.960955198669214e-05,
+      "loss": 0.3438,
+      "step": 14938
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9523925001171717e-05,
+      "loss": 0.4142,
+      "step": 14939
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9438465145288422e-05,
+      "loss": 0.9134,
+      "step": 14940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9353172596521618e-05,
+      "loss": 0.3995,
+      "step": 14941
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9268047532004007e-05,
+      "loss": 0.8528,
+      "step": 14942
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9183090128519678e-05,
+      "loss": 1.5228,
+      "step": 14943
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.909830056250531e-05,
+      "loss": 1.8487,
+      "step": 14944
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9013679010048223e-05,
+      "loss": 4.4692,
+      "step": 14945
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.892922564688766e-05,
+      "loss": 1.1073,
+      "step": 14946
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8844940648412702e-05,
+      "loss": 1.4786,
+      "step": 14947
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8760824189663627e-05,
+      "loss": 0.9736,
+      "step": 14948
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8676876445329884e-05,
+      "loss": 2.0402,
+      "step": 14949
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.859309758975134e-05,
+      "loss": 3.6237,
+      "step": 14950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8509487796916302e-05,
+      "loss": 2.0492,
+      "step": 14951
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.842604724046282e-05,
+      "loss": 0.2507,
+      "step": 14952
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8342776093676837e-05,
+      "loss": 1.1734,
+      "step": 14953
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.825967452949249e-05,
+      "loss": 1.7667,
+      "step": 14954
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8176742720492334e-05,
+      "loss": 3.4912,
+      "step": 14955
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8093980838905577e-05,
+      "loss": 3.442,
+      "step": 14956
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.801138905660934e-05,
+      "loss": 1.0425,
+      "step": 14957
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.792896754512686e-05,
+      "loss": 0.4662,
+      "step": 14958
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.784671647562781e-05,
+      "loss": 2.0374,
+      "step": 14959
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.776463601892845e-05,
+      "loss": 0.1674,
+      "step": 14960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.768272634548993e-05,
+      "loss": 3.9093,
+      "step": 14961
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.760098762541954e-05,
+      "loss": 0.9954,
+      "step": 14962
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7519420028468735e-05,
+      "loss": 0.8196,
+      "step": 14963
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.743802372403418e-05,
+      "loss": 0.2377,
+      "step": 14964
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.73567988811568e-05,
+      "loss": 0.9277,
+      "step": 14965
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.727574566852085e-05,
+      "loss": 3.3252,
+      "step": 14966
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7194864254454933e-05,
+      "loss": 3.2371,
+      "step": 14967
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7114154806930105e-05,
+      "loss": 0.5305,
+      "step": 14968
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.703361749356105e-05,
+      "loss": 2.7176,
+      "step": 14969
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6953252481604165e-05,
+      "loss": 2.5444,
+      "step": 14970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.687305993795886e-05,
+      "loss": 1.2219,
+      "step": 14971
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.679304002916553e-05,
+      "loss": 1.819,
+      "step": 14972
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6713192921406874e-05,
+      "loss": 1.3972,
+      "step": 14973
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6633518780505907e-05,
+      "loss": 2.3322,
+      "step": 14974
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.655401777192723e-05,
+      "loss": 3.0632,
+      "step": 14975
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.647469006077512e-05,
+      "loss": 3.0095,
+      "step": 14976
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6395535811794728e-05,
+      "loss": 2.3902,
+      "step": 14977
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6316555189370162e-05,
+      "loss": 2.4214,
+      "step": 14978
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6237748357525773e-05,
+      "loss": 3.7414,
+      "step": 14979
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.615911547992418e-05,
+      "loss": 0.8774,
+      "step": 14980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6080656719867525e-05,
+      "loss": 0.36,
+      "step": 14981
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.600237224029557e-05,
+      "loss": 2.5469,
+      "step": 14982
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5924262203786934e-05,
+      "loss": 0.1989,
+      "step": 14983
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.584632677255734e-05,
+      "loss": 2.2501,
+      "step": 14984
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5768566108459927e-05,
+      "loss": 4.3293,
+      "step": 14985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5690980372985464e-05,
+      "loss": 2.1708,
+      "step": 14986
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5613569727260656e-05,
+      "loss": 1.5984,
+      "step": 14987
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5536334332049362e-05,
+      "loss": 0.9871,
+      "step": 14988
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5459274347750886e-05,
+      "loss": 1.1254,
+      "step": 14989
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5382389934400286e-05,
+      "loss": 0.5032,
+      "step": 14990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5305681251668545e-05,
+      "loss": 0.2091,
+      "step": 14991
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5229148458860954e-05,
+      "loss": 0.1772,
+      "step": 14992
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5152791714918258e-05,
+      "loss": 3.6647,
+      "step": 14993
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5076611178414857e-05,
+      "loss": 0.6857,
+      "step": 14994
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5000607007559953e-05,
+      "loss": 2.2544,
+      "step": 14995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4924779360195728e-05,
+      "loss": 1.8048,
+      "step": 14996
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4849128393798317e-05,
+      "loss": 2.5387,
+      "step": 14997
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4773654265476966e-05,
+      "loss": 0.1343,
+      "step": 14998
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4698357131973061e-05,
+      "loss": 1.6002,
+      "step": 14999
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4623237149661207e-05,
+      "loss": 3.0823,
+      "step": 15000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454829447454732e-05,
+      "loss": 1.7011,
+      "step": 15001
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.447352926226987e-05,
+      "loss": 0.1419,
+      "step": 15002
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.439894166809802e-05,
+      "loss": 0.9552,
+      "step": 15003
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4324531846932799e-05,
+      "loss": 0.6833,
+      "step": 15004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4250299953305313e-05,
+      "loss": 0.7868,
+      "step": 15005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.417624614137789e-05,
+      "loss": 0.9522,
+      "step": 15006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.410237056494229e-05,
+      "loss": 0.8353,
+      "step": 15007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4028673377420854e-05,
+      "loss": 0.3578,
+      "step": 15008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3955154731864717e-05,
+      "loss": 2.5466,
+      "step": 15009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.388181478095496e-05,
+      "loss": 2.8488,
+      "step": 15010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3808653677000816e-05,
+      "loss": 1.2258,
+      "step": 15011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3735671571940844e-05,
+      "loss": 3.1818,
+      "step": 15012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3662868617341129e-05,
+      "loss": 2.8105,
+      "step": 15013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3590244964396414e-05,
+      "loss": 2.727,
+      "step": 15014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.351780076392838e-05,
+      "loss": 1.2533,
+      "step": 15015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.344553616638673e-05,
+      "loss": 2.4151,
+      "step": 15016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.337345132184763e-05,
+      "loss": 0.4589,
+      "step": 15017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3301546380013951e-05,
+      "loss": 0.0407,
+      "step": 15018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.32298214902155e-05,
+      "loss": 1.6999,
+      "step": 15019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3158276801407599e-05,
+      "loss": 0.7645,
+      "step": 15020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3086912462171342e-05,
+      "loss": 1.8524,
+      "step": 15021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.301572862071383e-05,
+      "loss": 2.9174,
+      "step": 15022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2944725424866566e-05,
+      "loss": 0.98,
+      "step": 15023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2873903022086652e-05,
+      "loss": 1.2322,
+      "step": 15024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2803261559454982e-05,
+      "loss": 1.4864,
+      "step": 15025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2732801183677411e-05,
+      "loss": 0.1752,
+      "step": 15026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.266252204108298e-05,
+      "loss": 2.8295,
+      "step": 15027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2592424277624925e-05,
+      "loss": 0.3831,
+      "step": 15028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2522508038879755e-05,
+      "loss": 1.0193,
+      "step": 15029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.245277347004652e-05,
+      "loss": 3.2147,
+      "step": 15030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.238322071594763e-05,
+      "loss": 3.107,
+      "step": 15031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2313849921027253e-05,
+      "loss": 0.1264,
+      "step": 15032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.224466122935235e-05,
+      "loss": 0.8765,
+      "step": 15033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2175654784611002e-05,
+      "loss": 1.0168,
+      "step": 15034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2106830730113494e-05,
+      "loss": 0.8552,
+      "step": 15035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2038189208790673e-05,
+      "loss": 2.6842,
+      "step": 15036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1969730363194953e-05,
+      "loss": 1.2215,
+      "step": 15037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1901454335498697e-05,
+      "loss": 0.5939,
+      "step": 15038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1833361267495313e-05,
+      "loss": 1.9121,
+      "step": 15039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.176545130059752e-05,
+      "loss": 3.3953,
+      "step": 15040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1697724575838497e-05,
+      "loss": 0.7405,
+      "step": 15041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1630181233870175e-05,
+      "loss": 0.7017,
+      "step": 15042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.156282141496432e-05,
+      "loss": 0.4603,
+      "step": 15043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.14956452590109e-05,
+      "loss": 3.7203,
+      "step": 15044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.142865290551911e-05,
+      "loss": 1.8018,
+      "step": 15045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1361844493615793e-05,
+      "loss": 1.0247,
+      "step": 15046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1295220162046427e-05,
+      "loss": 1.6712,
+      "step": 15047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1228780049173703e-05,
+      "loss": 0.7519,
+      "step": 15048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1162524292977772e-05,
+      "loss": 0.2587,
+      "step": 15049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1096453031056352e-05,
+      "loss": 2.5945,
+      "step": 15050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.1030566400623576e-05,
+      "loss": 3.9723,
+      "step": 15051
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.09648645385101e-05,
+      "loss": 0.3837,
+      "step": 15052
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0899347581163377e-05,
+      "loss": 1.2017,
+      "step": 15053
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.083401566464619e-05,
+      "loss": 1.4361,
+      "step": 15054
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0768868924637688e-05,
+      "loss": 1.1544,
+      "step": 15055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0703907496431809e-05,
+      "loss": 2.2577,
+      "step": 15056
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0639131514938294e-05,
+      "loss": 0.1429,
+      "step": 15057
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0574541114681103e-05,
+      "loss": 0.4985,
+      "step": 15058
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0510136429799422e-05,
+      "loss": 0.5632,
+      "step": 15059
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0445917594046128e-05,
+      "loss": 1.9639,
+      "step": 15060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0381884740788594e-05,
+      "loss": 0.1453,
+      "step": 15061
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0318038003007979e-05,
+      "loss": 1.3364,
+      "step": 15062
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0254377513298419e-05,
+      "loss": 1.6679,
+      "step": 15063
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0190903403867902e-05,
+      "loss": 0.8837,
+      "step": 15064
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0127615806536716e-05,
+      "loss": 0.174,
+      "step": 15065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0064514852738437e-05,
+      "loss": 0.3178,
+      "step": 15066
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0001600673518418e-05,
+      "loss": 0.4502,
+      "step": 15067
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.938873399534721e-06,
+      "loss": 0.123,
+      "step": 15068
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.876333161056628e-06,
+      "loss": 0.7795,
+      "step": 15069
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.813980087965647e-06,
+      "loss": 0.8248,
+      "step": 15070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.751814309753949e-06,
+      "loss": 2.6553,
+      "step": 15071
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.689835955525329e-06,
+      "loss": 0.0636,
+      "step": 15072
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.628045153993791e-06,
+      "loss": 2.1888,
+      "step": 15073
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.56644203348439e-06,
+      "loss": 2.3838,
+      "step": 15074
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.505026721931831e-06,
+      "loss": 1.3233,
+      "step": 15075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.44379934688141e-06,
+      "loss": 0.1092,
+      "step": 15076
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.38276003548747e-06,
+      "loss": 0.3951,
+      "step": 15077
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.321908914514432e-06,
+      "loss": 0.5198,
+      "step": 15078
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.261246110335441e-06,
+      "loss": 0.1182,
+      "step": 15079
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.200771748932524e-06,
+      "loss": 2.5953,
+      "step": 15080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.140485955896827e-06,
+      "loss": 1.7775,
+      "step": 15081
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.08038885642728e-06,
+      "loss": 4.4684,
+      "step": 15082
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.020480575331547e-06,
+      "loss": 1.2877,
+      "step": 15083
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.960761237024707e-06,
+      "loss": 0.7806,
+      "step": 15084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.90123096552945e-06,
+      "loss": 0.1229,
+      "step": 15085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.841889884476261e-06,
+      "loss": 1.7892,
+      "step": 15086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.78273811710213e-06,
+      "loss": 0.576,
+      "step": 15087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.723775786251499e-06,
+      "loss": 2.1434,
+      "step": 15088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.665003014374773e-06,
+      "loss": 1.2122,
+      "step": 15089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.606419923529297e-06,
+      "loss": 0.0948,
+      "step": 15090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.548026635377904e-06,
+      "loss": 0.1332,
+      "step": 15091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.48982327118969e-06,
+      "loss": 2.3648,
+      "step": 15092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.43180995183931e-06,
+      "loss": 0.7968,
+      "step": 15093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.373986797806343e-06,
+      "loss": 0.3829,
+      "step": 15094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.316353929175968e-06,
+      "loss": 3.3083,
+      "step": 15095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.258911465637653e-06,
+      "loss": 3.0458,
+      "step": 15096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.201659526485994e-06,
+      "loss": 1.4359,
+      "step": 15097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.14459823061935e-06,
+      "loss": 1.7737,
+      "step": 15098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.087727696540703e-06,
+      "loss": 1.6748,
+      "step": 15099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 8.031048042356348e-06,
+      "loss": 0.0769,
+      "step": 15100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.974559385776691e-06,
+      "loss": 0.1991,
+      "step": 15101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.918261844114937e-06,
+      "loss": 1.3197,
+      "step": 15102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.862155534287952e-06,
+      "loss": 0.3186,
+      "step": 15103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.806240572814893e-06,
+      "loss": 0.2461,
+      "step": 15104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.750517075818098e-06,
+      "loss": 1.5166,
+      "step": 15105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.694985159021705e-06,
+      "loss": 1.7333,
+      "step": 15106
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.639644937752577e-06,
+      "loss": 1.1595,
+      "step": 15107
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.584496526938878e-06,
+      "loss": 2.3182,
+      "step": 15108
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.529540041110994e-06,
+      "loss": 2.8275,
+      "step": 15109
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.474775594400185e-06,
+      "loss": 0.4217,
+      "step": 15110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.42020330053943e-06,
+      "loss": 0.2243,
+      "step": 15111
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.36582327286226e-06,
+      "loss": 1.6879,
+      "step": 15112
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.311635624302881e-06,
+      "loss": 3.9264,
+      "step": 15113
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.257640467396421e-06,
+      "loss": 1.1788,
+      "step": 15114
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.203837914277811e-06,
+      "loss": 0.2263,
+      "step": 15115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.1502280766820194e-06,
+      "loss": 0.7749,
+      "step": 15116
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.096811065944153e-06,
+      "loss": 2.9956,
+      "step": 15117
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.043586992998363e-06,
+      "loss": 0.8042,
+      "step": 15118
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.990555968378642e-06,
+      "loss": 3.661,
+      "step": 15119
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.937718102217517e-06,
+      "loss": 1.6986,
+      "step": 15120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.885073504246919e-06,
+      "loss": 2.2105,
+      "step": 15121
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.832622283796874e-06,
+      "loss": 2.7373,
+      "step": 15122
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.780364549796348e-06,
+      "loss": 0.2888,
+      "step": 15123
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.728300410771915e-06,
+      "loss": 2.4399,
+      "step": 15124
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.676429974848541e-06,
+      "loss": 2.7323,
+      "step": 15125
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.624753349748891e-06,
+      "loss": 3.1787,
+      "step": 15126
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.57327064279275e-06,
+      "loss": 0.2072,
+      "step": 15127
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.521981960897683e-06,
+      "loss": 0.1102,
+      "step": 15128
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.470887410577809e-06,
+      "loss": 1.1316,
+      "step": 15129
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.419987097944602e-06,
+      "loss": 2.659,
+      "step": 15130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.369281128705629e-06,
+      "loss": 0.1886,
+      "step": 15131
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.318769608165365e-06,
+      "loss": 2.0627,
+      "step": 15132
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.268452641223943e-06,
+      "loss": 0.2855,
+      "step": 15133
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.218330332377964e-06,
+      "loss": 4.2262,
+      "step": 15134
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.1684027857192965e-06,
+      "loss": 1.9748,
+      "step": 15135
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.118670104935786e-06,
+      "loss": 2.5617,
+      "step": 15136
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.0691323933101396e-06,
+      "loss": 0.1988,
+      "step": 15137
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.0197897537206064e-06,
+      "loss": 0.9503,
+      "step": 15138
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.970642288639861e-06,
+      "loss": 0.1967,
+      "step": 15139
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.921690100135724e-06,
+      "loss": 0.8503,
+      "step": 15140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.872933289869986e-06,
+      "loss": 2.5578,
+      "step": 15141
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.824371959099173e-06,
+      "loss": 1.1151,
+      "step": 15142
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.776006208673468e-06,
+      "loss": 2.1848,
+      "step": 15143
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.727836139036901e-06,
+      "loss": 1.0507,
+      "step": 15144
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.679861850227486e-06,
+      "loss": 0.3263,
+      "step": 15145
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.632083441876246e-06,
+      "loss": 0.8716,
+      "step": 15146
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.584501013207444e-06,
+      "loss": 0.0874,
+      "step": 15147
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.537114663038689e-06,
+      "loss": 0.8052,
+      "step": 15148
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.489924489779918e-06,
+      "loss": 1.5583,
+      "step": 15149
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.442930591434092e-06,
+      "loss": 3.3559,
+      "step": 15150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.3961330655961205e-06,
+      "loss": 2.0108,
+      "step": 15151
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.34953200945355e-06,
+      "loss": 0.2183,
+      "step": 15152
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.303127519785445e-06,
+      "loss": 2.2236,
+      "step": 15153
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.256919692963103e-06,
+      "loss": 0.153,
+      "step": 15154
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.210908624948929e-06,
+      "loss": 0.8618,
+      "step": 15155
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.165094411297089e-06,
+      "loss": 0.9826,
+      "step": 15156
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.1194771471528955e-06,
+      "loss": 1.8957,
+      "step": 15157
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.074056927252302e-06,
+      "loss": 1.243,
+      "step": 15158
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.028833845922531e-06,
+      "loss": 0.122,
+      "step": 15159
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.983807997080902e-06,
+      "loss": 1.0448,
+      "step": 15160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.938979474235617e-06,
+      "loss": 1.7209,
+      "step": 15161
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.894348370484614e-06,
+      "loss": 0.3311,
+      "step": 15162
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.849914778516284e-06,
+      "loss": 1.3203,
+      "step": 15163
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.805678790608381e-06,
+      "loss": 2.9757,
+      "step": 15164
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.76164049862875e-06,
+      "loss": 3.1028,
+      "step": 15165
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.717799994034211e-06,
+      "loss": 2.4196,
+      "step": 15166
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.6741573678712565e-06,
+      "loss": 2.9418,
+      "step": 15167
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.63071271077502e-06,
+      "loss": 0.2174,
+      "step": 15168
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.587466112969918e-06,
+      "loss": 0.16,
+      "step": 15169
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.5444176642686075e-06,
+      "loss": 0.1056,
+      "step": 15170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.501567454072675e-06,
+      "loss": 1.7141,
+      "step": 15171
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.458915571371602e-06,
+      "loss": 2.5651,
+      "step": 15172
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.4164621047433906e-06,
+      "loss": 3.0267,
+      "step": 15173
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.374207142353681e-06,
+      "loss": 2.508,
+      "step": 15174
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.332150771955845e-06,
+      "loss": 0.6615,
+      "step": 15175
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.290293080891161e-06,
+      "loss": 0.6643,
+      "step": 15176
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.248634156087971e-06,
+      "loss": 3.1719,
+      "step": 15177
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.207174084061849e-06,
+      "loss": 1.103,
+      "step": 15178
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.165912950915685e-06,
+      "loss": 1.6569,
+      "step": 15179
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.124850842338824e-06,
+      "loss": 0.8037,
+      "step": 15180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.083987843607673e-06,
+      "loss": 0.5145,
+      "step": 15181
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.0433240395847614e-06,
+      "loss": 3.1395,
+      "step": 15182
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.002859514719326e-06,
+      "loss": 3.6946,
+      "step": 15183
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.962594353046323e-06,
+      "loss": 0.6612,
+      "step": 15184
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.922528638187117e-06,
+      "loss": 2.7051,
+      "step": 15185
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.882662453348463e-06,
+      "loss": 2.5075,
+      "step": 15186
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.842995881323108e-06,
+      "loss": 2.6511,
+      "step": 15187
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8035290044888814e-06,
+      "loss": 0.3708,
+      "step": 15188
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7642619048091964e-06,
+      "loss": 1.3777,
+      "step": 15189
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.725194663832543e-06,
+      "loss": 1.2899,
+      "step": 15190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6863273626920993e-06,
+      "loss": 0.767,
+      "step": 15191
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6476600821061634e-06,
+      "loss": 2.3524,
+      "step": 15192
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6091929023772896e-06,
+      "loss": 0.172,
+      "step": 15193
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5709259033928298e-06,
+      "loss": 0.5721,
+      "step": 15194
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.532859164624047e-06,
+      "loss": 0.6096,
+      "step": 15195
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.494992765126659e-06,
+      "loss": 0.7466,
+      "step": 15196
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4573267835399824e-06,
+      "loss": 0.7027,
+      "step": 15197
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.419861298087479e-06,
+      "loss": 2.0016,
+      "step": 15198
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.3825963865758315e-06,
+      "loss": 2.0928,
+      "step": 15199
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.3455321263955896e-06,
+      "loss": 2.6784,
+      "step": 15200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.3086685945202036e-06,
+      "loss": 3.0787,
+      "step": 15201
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.2720058675066222e-06,
+      "loss": 1.2547,
+      "step": 15202
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.2355440214944187e-06,
+      "loss": 0.8059,
+      "step": 15203
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.1992831322063434e-06,
+      "loss": 1.3043,
+      "step": 15204
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.1632232749475355e-06,
+      "loss": 0.5858,
+      "step": 15205
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.127364524605636e-06,
+      "loss": 1.4464,
+      "step": 15206
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0917069556509083e-06,
+      "loss": 2.1849,
+      "step": 15207
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0562506421354496e-06,
+      "loss": 0.5382,
+      "step": 15208
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.020995657693748e-06,
+      "loss": 0.3526,
+      "step": 15209
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9859420755419367e-06,
+      "loss": 0.9623,
+      "step": 15210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9510899684778494e-06,
+      "loss": 0.583,
+      "step": 15211
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9164394088811887e-06,
+      "loss": 0.1656,
+      "step": 15212
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8819904687127586e-06,
+      "loss": 2.7732,
+      "step": 15213
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8477432195149423e-06,
+      "loss": 1.1825,
+      "step": 15214
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.813697732410969e-06,
+      "loss": 0.4164,
+      "step": 15215
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7798540781053704e-06,
+      "loss": 0.2784,
+      "step": 15216
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7462123268831798e-06,
+      "loss": 0.5362,
+      "step": 15217
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.712772548610476e-06,
+      "loss": 1.5411,
+      "step": 15218
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.679534812733542e-06,
+      "loss": 5.3375,
+      "step": 15219
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.6464991882793167e-06,
+      "loss": 2.9017,
+      "step": 15220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.6136657438550095e-06,
+      "loss": 1.6867,
+      "step": 15221
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.581034547647698e-06,
+      "loss": 0.2011,
+      "step": 15222
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.5486056674247637e-06,
+      "loss": 2.4136,
+      "step": 15223
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.5163791705331117e-06,
+      "loss": 1.6721,
+      "step": 15224
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.484355123899651e-06,
+      "loss": 0.1557,
+      "step": 15225
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.452533594030548e-06,
+      "loss": 0.1343,
+      "step": 15226
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.420914647011685e-06,
+      "loss": 2.7626,
+      "step": 15227
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.389498348507935e-06,
+      "loss": 1.4596,
+      "step": 15228
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.3582847637636187e-06,
+      "loss": 0.4697,
+      "step": 15229
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.327273957601761e-06,
+      "loss": 4.6622,
+      "step": 15230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.296465994424568e-06,
+      "loss": 1.7347,
+      "step": 15231
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.265860938212705e-06,
+      "loss": 0.9295,
+      "step": 15232
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.235458852525718e-06,
+      "loss": 1.9877,
+      "step": 15233
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.205259800501347e-06,
+      "loss": 0.7376,
+      "step": 15234
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.175263844855979e-06,
+      "loss": 0.7219,
+      "step": 15235
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.14547104788394e-06,
+      "loss": 0.3856,
+      "step": 15236
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.115881471457892e-06,
+      "loss": 1.2326,
+      "step": 15237
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0864951770282693e-06,
+      "loss": 1.288,
+      "step": 15238
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0573122256233425e-06,
+      "loss": 1.1618,
+      "step": 15239
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0283326778492873e-06,
+      "loss": 0.9525,
+      "step": 15240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.999556593889651e-06,
+      "loss": 1.4045,
+      "step": 15241
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.97098403350543e-06,
+      "loss": 2.097,
+      "step": 15242
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9426150560351796e-06,
+      "loss": 1.4874,
+      "step": 15243
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.914449720394351e-06,
+      "loss": 1.1207,
+      "step": 15244
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.8864880850757882e-06,
+      "loss": 0.1961,
+      "step": 15245
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.8587302081489954e-06,
+      "loss": 2.2141,
+      "step": 15246
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.831176147260627e-06,
+      "loss": 0.8725,
+      "step": 15247
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.8038259596337537e-06,
+      "loss": 1.7496,
+      "step": 15248
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.77667970206834e-06,
+      "loss": 0.8603,
+      "step": 15249
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7497374309405568e-06,
+      "loss": 0.1719,
+      "step": 15250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7229992022032237e-06,
+      "loss": 0.5792,
+      "step": 15251
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6964650713851338e-06,
+      "loss": 0.7479,
+      "step": 15252
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6701350935914296e-06,
+      "loss": 0.1495,
+      "step": 15253
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6440093235032594e-06,
+      "loss": 0.5086,
+      "step": 15254
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6180878153774893e-06,
+      "loss": 4.1938,
+      "step": 15255
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.592370623047068e-06,
+      "loss": 2.9684,
+      "step": 15256
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5668577999203294e-06,
+      "loss": 0.816,
+      "step": 15257
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5415493989814567e-06,
+      "loss": 2.6684,
+      "step": 15258
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5164454727898514e-06,
+      "loss": 0.2219,
+      "step": 15259
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4915460734805097e-06,
+      "loss": 0.4255,
+      "step": 15260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.466851252763435e-06,
+      "loss": 2.9918,
+      "step": 15261
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4423610619240025e-06,
+      "loss": 1.3553,
+      "step": 15262
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4180755518223955e-06,
+      "loss": 0.339,
+      "step": 15263
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3939947728939251e-06,
+      "loss": 0.4415,
+      "step": 15264
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3701187751485434e-06,
+      "loss": 2.7116,
+      "step": 15265
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3464476081711198e-06,
+      "loss": 2.0563,
+      "step": 15266
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3229813211208974e-06,
+      "loss": 0.889,
+      "step": 15267
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2997199627318934e-06,
+      "loss": 0.103,
+      "step": 15268
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.276663581312354e-06,
+      "loss": 0.9182,
+      "step": 15269
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2538122247448325e-06,
+      "loss": 1.9185,
+      "step": 15270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2311659404862563e-06,
+      "loss": 3.5824,
+      "step": 15271
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2087247755675158e-06,
+      "loss": 0.1595,
+      "step": 15272
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.1864887765934973e-06,
+      "loss": 0.6703,
+      "step": 15273
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.1644579897431952e-06,
+      "loss": 2.4183,
+      "step": 15274
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.1426324607691996e-06,
+      "loss": 2.2603,
+      "step": 15275
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.1210122349980423e-06,
+      "loss": 0.555,
+      "step": 15276
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0995973573296625e-06,
+      "loss": 3.2393,
+      "step": 15277
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0783878722377738e-06,
+      "loss": 2.0197,
+      "step": 15278
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.057383823769309e-06,
+      "loss": 1.4176,
+      "step": 15279
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0365852555447975e-06,
+      "loss": 1.7805,
+      "step": 15280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.015992210757799e-06,
+      "loss": 0.1774,
+      "step": 15281
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.956047321752592e-07,
+      "loss": 1.3977,
+      "step": 15282
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.754228621370209e-07,
+      "loss": 0.3079,
+      "step": 15283
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.55446642556057e-07,
+      "loss": 1.9815,
+      "step": 15284
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.356761149182825e-07,
+      "loss": 0.513,
+      "step": 15285
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.161113202822646e-07,
+      "loss": 1.5673,
+      "step": 15286
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.967522992795352e-07,
+      "loss": 4.4195,
+      "step": 15287
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.775990921140453e-07,
+      "loss": 0.15,
+      "step": 15288
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.586517385625548e-07,
+      "loss": 1.6801,
+      "step": 15289
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.399102779741097e-07,
+      "loss": 0.1075,
+      "step": 15290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.213747492703538e-07,
+      "loss": 0.0994,
+      "step": 15291
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 8.030451909450731e-07,
+      "loss": 1.3844,
+      "step": 15292
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.849216410644733e-07,
+      "loss": 0.079,
+      "step": 15293
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.670041372667691e-07,
+      "loss": 0.1689,
+      "step": 15294
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.492927167624287e-07,
+      "loss": 1.025,
+      "step": 15295
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.317874163337846e-07,
+      "loss": 0.7535,
+      "step": 15296
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 7.144882723352453e-07,
+      "loss": 0.6743,
+      "step": 15297
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.973953206929285e-07,
+      "loss": 1.7754,
+      "step": 15298
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.805085969049052e-07,
+      "loss": 1.761,
+      "step": 15299
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.63828136040856e-07,
+      "loss": 0.5442,
+      "step": 15300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.473539727420597e-07,
+      "loss": 4.5963,
+      "step": 15301
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.310861412215152e-07,
+      "loss": 0.6838,
+      "step": 15302
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.150246752635535e-07,
+      "loss": 1.5277,
+      "step": 15303
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.991696082241038e-07,
+      "loss": 0.2076,
+      "step": 15304
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.835209730303159e-07,
+      "loss": 2.9778,
+      "step": 15305
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.680788021806382e-07,
+      "loss": 2.1614,
+      "step": 15306
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.528431277448509e-07,
+      "loss": 0.1218,
+      "step": 15307
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.378139813637551e-07,
+      "loss": 0.1384,
+      "step": 15308
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.229913942493836e-07,
+      "loss": 2.3653,
+      "step": 15309
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.08375397184635e-07,
+      "loss": 0.2154,
+      "step": 15310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.939660205235397e-07,
+      "loss": 2.4186,
+      "step": 15311
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.797632941908603e-07,
+      "loss": 1.0045,
+      "step": 15312
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.6576724768233606e-07,
+      "loss": 0.8278,
+      "step": 15313
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.5197791006437174e-07,
+      "loss": 0.1972,
+      "step": 15314
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.3839530997419333e-07,
+      "loss": 0.2837,
+      "step": 15315
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.250194756195702e-07,
+      "loss": 0.2591,
+      "step": 15316
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.118504347789376e-07,
+      "loss": 2.2323,
+      "step": 15317
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.9888821480130733e-07,
+      "loss": 0.9005,
+      "step": 15318
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.861328426060462e-07,
+      "loss": 1.8609,
+      "step": 15319
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.735843446830978e-07,
+      "loss": 0.14,
+      "step": 15320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6124274709262717e-07,
+      "loss": 0.9748,
+      "step": 15321
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.4910807546526534e-07,
+      "loss": 0.1587,
+      "step": 15322
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.371803550017649e-07,
+      "loss": 0.8723,
+      "step": 15323
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.2545961047322213e-07,
+      "loss": 1.7373,
+      "step": 15324
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.1394586622077725e-07,
+      "loss": 0.4914,
+      "step": 15325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.02639146155792e-07,
+      "loss": 3.3458,
+      "step": 15326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.9153947375957227e-07,
+      "loss": 1.643,
+      "step": 15327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.8064687208354534e-07,
+      "loss": 0.1241,
+      "step": 15328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6996136374900507e-07,
+      "loss": 0.1545,
+      "step": 15329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.594829709472446e-07,
+      "loss": 1.6107,
+      "step": 15330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4921171543937914e-07,
+      "loss": 3.0806,
+      "step": 15331
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.391476185563235e-07,
+      "loss": 3.5991,
+      "step": 15332
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2929070119885876e-07,
+      "loss": 0.2286,
+      "step": 15333
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.1964098383742137e-07,
+      "loss": 0.6142,
+      "step": 15334
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.1019848651220308e-07,
+      "loss": 0.7735,
+      "step": 15335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.0096322883300656e-07,
+      "loss": 0.148,
+      "step": 15336
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.919352299792454e-07,
+      "loss": 0.5949,
+      "step": 15337
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8311450869995528e-07,
+      "loss": 1.2897,
+      "step": 15338
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7450108331363845e-07,
+      "loss": 0.5179,
+      "step": 15339
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6609497170836373e-07,
+      "loss": 1.6039,
+      "step": 15340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5789619134157774e-07,
+      "loss": 2.2535,
+      "step": 15341
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4990475924024916e-07,
+      "loss": 0.2024,
+      "step": 15342
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4212069200063573e-07,
+      "loss": 2.1682,
+      "step": 15343
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.345440057884173e-07,
+      "loss": 0.344,
+      "step": 15344
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2717471633854062e-07,
+      "loss": 0.9361,
+      "step": 15345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.200128389552968e-07,
+      "loss": 0.1655,
+      "step": 15346
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.13058388512155e-07,
+      "loss": 0.6029,
+      "step": 15347
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0631137945189551e-07,
+      "loss": 1.9872,
+      "step": 15348
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.977182578643219e-08,
+      "loss": 0.9213,
+      "step": 15349
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.343974109684572e-08,
+      "loss": 1.0969,
+      "step": 15350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.731513853340589e-08,
+      "loss": 1.2006,
+      "step": 15351
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.139803081540498e-08,
+      "loss": 0.6556,
+      "step": 15352
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.568843023130213e-08,
+      "loss": 0.1269,
+      "step": 15353
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.018634863855677e-08,
+      "loss": 3.3008,
+      "step": 15354
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.489179746370644e-08,
+      "loss": 1.0886,
+      "step": 15355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.980478770223341e-08,
+      "loss": 2.3137,
+      "step": 15356
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.492532991868693e-08,
+      "loss": 1.1298,
+      "step": 15357
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.0253434246505526e-08,
+      "loss": 0.7721,
+      "step": 15358
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.578911038813916e-08,
+      "loss": 1.059,
+      "step": 15359
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.153236761488266e-08,
+      "loss": 0.8973,
+      "step": 15360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.7483214767009e-08,
+      "loss": 4.8251,
+      "step": 15361
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.364166025362492e-08,
+      "loss": 0.972,
+      "step": 15362
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.0007712052737556e-08,
+      "loss": 2.4904,
+      "step": 15363
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.658137771118785e-08,
+      "loss": 2.8488,
+      "step": 15364
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.3362664344639407e-08,
+      "loss": 1.6573,
+      "step": 15365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.0351578637622938e-08,
+      "loss": 2.6824,
+      "step": 15366
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7548126843425215e-08,
+      "loss": 2.1029,
+      "step": 15367
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4952314784144607e-08,
+      "loss": 0.6562,
+      "step": 15368
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2564147850691044e-08,
+      "loss": 3.6439,
+      "step": 15369
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0383631002686134e-08,
+      "loss": 0.1383,
+      "step": 15370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.41076876856306e-09,
+      "loss": 0.3846,
+      "step": 15371
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.645565245488872e-09,
+      "loss": 2.6823,
+      "step": 15372
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.088024099353383e-09,
+      "loss": 2.9951,
+      "step": 15373
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.738148564813582e-09,
+      "loss": 0.9893,
+      "step": 15374
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.595941445238115e-09,
+      "loss": 0.2377,
+      "step": 15375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6614051127072928e-09,
+      "loss": 3.1969,
+      "step": 15376
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.345415080463936e-10,
+      "loss": 2.2237,
+      "step": 15377
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.1535214075905106e-10,
+      "loss": 0.6694,
+      "step": 15378
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.038380891049684e-10,
+      "loss": 1.5961,
+      "step": 15379
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0,
+      "loss": 1.0531,
+      "step": 15380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.038380891049684e-10,
+      "loss": 0.5145,
+      "step": 15381
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.1535214075905106e-10,
+      "loss": 0.3942,
+      "step": 15382
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.345415080463936e-10,
+      "loss": 1.7068,
+      "step": 15383
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6614051127072928e-09,
+      "loss": 1.4246,
+      "step": 15384
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5959414452270124e-09,
+      "loss": 4.1944,
+      "step": 15385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.738148564813582e-09,
+      "loss": 0.3644,
+      "step": 15386
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.088024099353383e-09,
+      "loss": 0.2271,
+      "step": 15387
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.64556524547777e-09,
+      "loss": 1.0494,
+      "step": 15388
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.41076876856306e-09,
+      "loss": 1.043,
+      "step": 15389
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0383631002686134e-08,
+      "loss": 2.5258,
+      "step": 15390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2564147850691044e-08,
+      "loss": 0.1735,
+      "step": 15391
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4952314784144607e-08,
+      "loss": 0.1031,
+      "step": 15392
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7548126843425215e-08,
+      "loss": 0.6449,
+      "step": 15393
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.0351578637611833e-08,
+      "loss": 0.1455,
+      "step": 15394
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.3362664344639407e-08,
+      "loss": 1.2605,
+      "step": 15395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6581377711176747e-08,
+      "loss": 1.1267,
+      "step": 15396
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.000771205272646e-08,
+      "loss": 0.5277,
+      "step": 15397
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3641660253613816e-08,
+      "loss": 0.2513,
+      "step": 15398
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.74832147669979e-08,
+      "loss": 2.1982,
+      "step": 15399
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.153236761487156e-08,
+      "loss": 1.5321,
+      "step": 15400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.5789110388128055e-08,
+      "loss": 3.0443,
+      "step": 15401
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.0253434246505526e-08,
+      "loss": 2.8588,
+      "step": 15402
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.492532991868693e-08,
+      "loss": 0.3436,
+      "step": 15403
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.980478770223341e-08,
+      "loss": 0.0558,
+      "step": 15404
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.489179746369534e-08,
+      "loss": 3.3379,
+      "step": 15405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.018634863855677e-08,
+      "loss": 0.8847,
+      "step": 15406
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.568843023130213e-08,
+      "loss": 0.148,
+      "step": 15407
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.139803081539387e-08,
+      "loss": 1.6078,
+      "step": 15408
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.731513853339479e-08,
+      "loss": 1.7132,
+      "step": 15409
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.343974109684572e-08,
+      "loss": 1.6417,
+      "step": 15410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.977182578642108e-08,
+      "loss": 1.0098,
+      "step": 15411
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.063113794518844e-07,
+      "loss": 0.3457,
+      "step": 15412
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.13058388512155e-07,
+      "loss": 2.0261,
+      "step": 15413
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.200128389552857e-07,
+      "loss": 0.1111,
+      "step": 15414
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.271747163385295e-07,
+      "loss": 1.3725,
+      "step": 15415
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.345440057884062e-07,
+      "loss": 0.3084,
+      "step": 15416
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.421206920006246e-07,
+      "loss": 0.2565,
+      "step": 15417
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4990475924023807e-07,
+      "loss": 1.5657,
+      "step": 15418
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.5789619134156662e-07,
+      "loss": 0.776,
+      "step": 15419
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6609497170835264e-07,
+      "loss": 1.6637,
+      "step": 15420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7450108331362736e-07,
+      "loss": 2.4895,
+      "step": 15421
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8311450869994416e-07,
+      "loss": 1.9273,
+      "step": 15422
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.919352299792343e-07,
+      "loss": 0.0935,
+      "step": 15423
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.0096322883299544e-07,
+      "loss": 1.4082,
+      "step": 15424
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.1019848651219199e-07,
+      "loss": 0.6698,
+      "step": 15425
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.1964098383741027e-07,
+      "loss": 3.2463,
+      "step": 15426
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2929070119884767e-07,
+      "loss": 0.6338,
+      "step": 15427
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.391476185563124e-07,
+      "loss": 2.9398,
+      "step": 15428
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.492117154393681e-07,
+      "loss": 1.5993,
+      "step": 15429
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.594829709472335e-07,
+      "loss": 0.5859,
+      "step": 15430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6996136374899395e-07,
+      "loss": 0.1749,
+      "step": 15431
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.8064687208353423e-07,
+      "loss": 0.5174,
+      "step": 15432
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.9153947375956115e-07,
+      "loss": 0.8383,
+      "step": 15433
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.0263914615578094e-07,
+      "loss": 0.612,
+      "step": 15434
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.139458662207662e-07,
+      "loss": 1.2004,
+      "step": 15435
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2545961047321107e-07,
+      "loss": 2.6235,
+      "step": 15436
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.371803550017538e-07,
+      "loss": 0.8791,
+      "step": 15437
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4910807546524316e-07,
+      "loss": 0.0943,
+      "step": 15438
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.6124274709261605e-07,
+      "loss": 0.13,
+      "step": 15439
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.7358434468307555e-07,
+      "loss": 1.0327,
+      "step": 15440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.861328426060351e-07,
+      "loss": 1.5015,
+      "step": 15441
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.988882148012962e-07,
+      "loss": 1.588,
+      "step": 15442
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.118504347789265e-07,
+      "loss": 2.4722,
+      "step": 15443
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.2501947561954804e-07,
+      "loss": 0.3956,
+      "step": 15444
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.383953099741822e-07,
+      "loss": 2.2206,
+      "step": 15445
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.519779100643606e-07,
+      "loss": 0.1877,
+      "step": 15446
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.6576724768232494e-07,
+      "loss": 2.4673,
+      "step": 15447
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.797632941908381e-07,
+      "loss": 0.8122,
+      "step": 15448
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.939660205235174e-07,
+      "loss": 1.6569,
+      "step": 15449
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.083753971846239e-07,
+      "loss": 0.4694,
+      "step": 15450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.229913942493613e-07,
+      "loss": 0.6083,
+      "step": 15451
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.37813981363744e-07,
+      "loss": 1.566,
+      "step": 15452
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.528431277448288e-07,
+      "loss": 2.0743,
+      "step": 15453
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.68078802180616e-07,
+      "loss": 3.1,
+      "step": 15454
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.835209730302938e-07,
+      "loss": 2.5034,
+      "step": 15455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.991696082240816e-07,
+      "loss": 1.9332,
+      "step": 15456
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.150246752635424e-07,
+      "loss": 0.136,
+      "step": 15457
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.31086141221493e-07,
+      "loss": 3.4538,
+      "step": 15458
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.473539727420374e-07,
+      "loss": 1.6789,
+      "step": 15459
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.638281360408339e-07,
+      "loss": 0.832,
+      "step": 15460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.805085969048941e-07,
+      "loss": 1.8932,
+      "step": 15461
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.973953206929063e-07,
+      "loss": 0.1343,
+      "step": 15462
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.14488272335223e-07,
+      "loss": 1.7981,
+      "step": 15463
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.317874163337623e-07,
+      "loss": 1.8225,
+      "step": 15464
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.492927167624175e-07,
+      "loss": 3.4446,
+      "step": 15465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.670041372667469e-07,
+      "loss": 0.1996,
+      "step": 15466
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 7.849216410644511e-07,
+      "loss": 0.7083,
+      "step": 15467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.03045190945051e-07,
+      "loss": 2.6375,
+      "step": 15468
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.213747492703316e-07,
+      "loss": 0.723,
+      "step": 15469
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.399102779740875e-07,
+      "loss": 0.7683,
+      "step": 15470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.586517385625326e-07,
+      "loss": 5.3282,
+      "step": 15471
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.775990921140231e-07,
+      "loss": 2.6788,
+      "step": 15472
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 8.967522992795019e-07,
+      "loss": 3.9671,
+      "step": 15473
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.161113202822425e-07,
+      "loss": 0.1238,
+      "step": 15474
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.356761149182602e-07,
+      "loss": 4.8297,
+      "step": 15475
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.55446642556035e-07,
+      "loss": 1.6318,
+      "step": 15476
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.754228621369988e-07,
+      "loss": 0.438,
+      "step": 15477
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.95604732175237e-07,
+      "loss": 0.074,
+      "step": 15478
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0159922107577769e-06,
+      "loss": 0.8939,
+      "step": 15479
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0365852555447753e-06,
+      "loss": 0.6534,
+      "step": 15480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0573838237692869e-06,
+      "loss": 0.124,
+      "step": 15481
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0783878722377405e-06,
+      "loss": 1.5294,
+      "step": 15482
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0995973573296293e-06,
+      "loss": 1.2049,
+      "step": 15483
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.121012234998009e-06,
+      "loss": 4.2982,
+      "step": 15484
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1426324607691774e-06,
+      "loss": 1.4312,
+      "step": 15485
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.164457989743173e-06,
+      "loss": 1.095,
+      "step": 15486
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.186488776593475e-06,
+      "loss": 1.4282,
+      "step": 15487
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2087247755674824e-06,
+      "loss": 0.6507,
+      "step": 15488
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.231165940486223e-06,
+      "loss": 1.4165,
+      "step": 15489
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2538122247447993e-06,
+      "loss": 0.6674,
+      "step": 15490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2766635813123317e-06,
+      "loss": 1.4886,
+      "step": 15491
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2997199627318712e-06,
+      "loss": 0.7746,
+      "step": 15492
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3229813211208752e-06,
+      "loss": 1.9057,
+      "step": 15493
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3464476081710863e-06,
+      "loss": 2.1041,
+      "step": 15494
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3701187751485211e-06,
+      "loss": 3.0529,
+      "step": 15495
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3939947728939029e-06,
+      "loss": 1.6139,
+      "step": 15496
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.418075551822362e-06,
+      "loss": 0.9071,
+      "step": 15497
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4423610619239802e-06,
+      "loss": 2.9957,
+      "step": 15498
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4668512527634127e-06,
+      "loss": 0.1502,
+      "step": 15499
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4915460734804877e-06,
+      "loss": 3.9125,
+      "step": 15500
+    }
+  ],
+  "max_steps": 18354,
+  "num_train_epochs": 1,
+  "total_flos": 2.159331508224e+16,
+  "trial_name": null,
+  "trial_params": null
+}