diff --git "a/checkpoint-8477/trainer_state.json" "b/checkpoint-8477/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-8477/trainer_state.json"
@@ -0,0 +1,50891 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9625278509216124,
+  "eval_steps": 500,
+  "global_step": 8477,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1302,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.2193,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1633,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1995,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1775,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1753,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.2371,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.2672,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.2261,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1937,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1238,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1763,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1434,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.2315,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.2097,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1639,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.1884,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-05,
+      "loss": 1.1501,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-05,
+      "loss": 1.2154,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-05,
+      "loss": 1.133,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012,
+      "loss": 1.0502,
+      "step": 21
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00012,
+      "loss": 1.1904,
+      "step": 22
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00016,
+      "loss": 1.1193,
+      "step": 23
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002,
+      "loss": 1.0192,
+      "step": 24
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999999834562185,
+      "loss": 0.9811,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999999338248756,
+      "loss": 0.9311,
+      "step": 26
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999998511059717,
+      "loss": 1.0179,
+      "step": 27
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999997352995104,
+      "loss": 0.9642,
+      "step": 28
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999995864054953,
+      "loss": 1.0291,
+      "step": 29
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999999404423931,
+      "loss": 0.912,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999999189354824,
+      "loss": 1.0022,
+      "step": 31
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999998941198181,
+      "loss": 1.0605,
+      "step": 32
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999998659954011,
+      "loss": 0.8987,
+      "step": 33
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999983456223222,
+      "loss": 0.9043,
+      "step": 34
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999979982031257,
+      "loss": 0.8478,
+      "step": 35
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999997617696433,
+      "loss": 0.9314,
+      "step": 36
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999972041022564,
+      "loss": 0.8786,
+      "step": 37
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000199999675742061,
+      "loss": 0.9615,
+      "step": 38
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999962776515078,
+      "loss": 0.9248,
+      "step": 39
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999957647949662,
+      "loss": 0.8965,
+      "step": 40
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999952188510026,
+      "loss": 0.9037,
+      "step": 41
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999946398196344,
+      "loss": 0.9219,
+      "step": 42
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999940277008808,
+      "loss": 0.9147,
+      "step": 43
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999933824947625,
+      "loss": 0.8407,
+      "step": 44
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999927042013002,
+      "loss": 0.9245,
+      "step": 45
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999991992820517,
+      "loss": 0.9034,
+      "step": 46
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999912483524362,
+      "loss": 0.9572,
+      "step": 47
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999904707970824,
+      "loss": 0.8646,
+      "step": 48
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999896601544809,
+      "loss": 0.8362,
+      "step": 49
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999988816424659,
+      "loss": 0.8362,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999879396076452,
+      "loss": 0.8844,
+      "step": 51
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999870297034673,
+      "loss": 0.9193,
+      "step": 52
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999986086712156,
+      "loss": 0.8375,
+      "step": 53
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999851106337427,
+      "loss": 0.7938,
+      "step": 54
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999841014682593,
+      "loss": 0.8335,
+      "step": 55
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999830592157394,
+      "loss": 0.8737,
+      "step": 56
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999819838762174,
+      "loss": 0.8544,
+      "step": 57
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999808754497288,
+      "loss": 0.8236,
+      "step": 58
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999797339363104,
+      "loss": 0.8693,
+      "step": 59
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999785593360003,
+      "loss": 0.8374,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999773516488366,
+      "loss": 0.8425,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999761108748597,
+      "loss": 0.8295,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0001999974837014111,
+      "loss": 0.9197,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019999735300666322,
+      "loss": 0.8137,
+      "step": 64
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999721900324665,
+      "loss": 0.8066,
+      "step": 65
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999708169116587,
+      "loss": 0.8622,
+      "step": 66
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999694107042534,
+      "loss": 0.9321,
+      "step": 67
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999679714102978,
+      "loss": 0.8322,
+      "step": 68
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999664990298394,
+      "loss": 0.8548,
+      "step": 69
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999964993562927,
+      "loss": 0.9037,
+      "step": 70
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000199996345500961,
+      "loss": 0.843,
+      "step": 71
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000199996188336994,
+      "loss": 0.7413,
+      "step": 72
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999602786439685,
+      "loss": 0.865,
+      "step": 73
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999586408317484,
+      "loss": 0.8781,
+      "step": 74
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999569699333348,
+      "loss": 0.8298,
+      "step": 75
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999552659487818,
+      "loss": 0.8432,
+      "step": 76
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999535288781463,
+      "loss": 0.8408,
+      "step": 77
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999951758721486,
+      "loss": 0.7852,
+      "step": 78
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999499554788593,
+      "loss": 0.8581,
+      "step": 79
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999948119150326,
+      "loss": 0.8628,
+      "step": 80
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999462497359466,
+      "loss": 0.8661,
+      "step": 81
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999944347235783,
+      "loss": 0.831,
+      "step": 82
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999424116498983,
+      "loss": 0.9166,
+      "step": 83
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999404429783566,
+      "loss": 0.8071,
+      "step": 84
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999384412212228,
+      "loss": 0.8683,
+      "step": 85
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999364063785633,
+      "loss": 0.7978,
+      "step": 86
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999343384504453,
+      "loss": 0.8396,
+      "step": 87
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999322374369372,
+      "loss": 0.8789,
+      "step": 88
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999301033381085,
+      "loss": 0.9197,
+      "step": 89
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999279361540301,
+      "loss": 0.8511,
+      "step": 90
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999257358847735,
+      "loss": 0.8512,
+      "step": 91
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999235025304116,
+      "loss": 0.8125,
+      "step": 92
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999212360910181,
+      "loss": 0.7839,
+      "step": 93
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999189365666685,
+      "loss": 0.7678,
+      "step": 94
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999916603957438,
+      "loss": 0.8549,
+      "step": 95
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999142382634045,
+      "loss": 0.886,
+      "step": 96
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999911839484646,
+      "loss": 0.8455,
+      "step": 97
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999909407621242,
+      "loss": 0.7946,
+      "step": 98
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999906942673273,
+      "loss": 0.7877,
+      "step": 99
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000199990444464082,
+      "loss": 0.7894,
+      "step": 100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019999019135239664,
+      "loss": 0.8507,
+      "step": 101
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019998993493227957,
+      "loss": 0.8617,
+      "step": 102
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019998967520373927,
+      "loss": 0.8381,
+      "step": 103
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019998941216678432,
+      "loss": 0.8568,
+      "step": 104
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019998914582142346,
+      "loss": 0.8431,
+      "step": 105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019998887616766545,
+      "loss": 0.7984,
+      "step": 106
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001999886032055193,
+      "loss": 0.8666,
+      "step": 107
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999883269349939,
+      "loss": 0.8023,
+      "step": 108
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998804735609852,
+      "loss": 0.8323,
+      "step": 109
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998776446884236,
+      "loss": 0.8321,
+      "step": 110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998747827323479,
+      "loss": 0.8164,
+      "step": 111
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998718876928527,
+      "loss": 0.8613,
+      "step": 112
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998689595700336,
+      "loss": 0.7596,
+      "step": 113
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998659983639882,
+      "loss": 0.8491,
+      "step": 114
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998630040748137,
+      "loss": 0.8224,
+      "step": 115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998599767026093,
+      "loss": 0.7451,
+      "step": 116
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998569162474756,
+      "loss": 0.731,
+      "step": 117
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998538227095135,
+      "loss": 0.854,
+      "step": 118
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998506960888256,
+      "loss": 0.8263,
+      "step": 119
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998475363855152,
+      "loss": 0.8542,
+      "step": 120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998443435996868,
+      "loss": 0.8027,
+      "step": 121
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999841117731446,
+      "loss": 0.9464,
+      "step": 122
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998378587808997,
+      "loss": 0.8516,
+      "step": 123
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998345667481557,
+      "loss": 0.8446,
+      "step": 124
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998312416333227,
+      "loss": 0.8324,
+      "step": 125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998278834365115,
+      "loss": 0.8817,
+      "step": 126
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999824492157832,
+      "loss": 0.8351,
+      "step": 127
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998210677973972,
+      "loss": 0.8708,
+      "step": 128
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998176103553205,
+      "loss": 0.7821,
+      "step": 129
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998141198317154,
+      "loss": 0.7743,
+      "step": 130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998105962266987,
+      "loss": 0.8352,
+      "step": 131
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999807039540386,
+      "loss": 0.8127,
+      "step": 132
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998034497728954,
+      "loss": 0.7933,
+      "step": 133
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997998269243456,
+      "loss": 0.7829,
+      "step": 134
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997961709948564,
+      "loss": 0.834,
+      "step": 135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997924819845488,
+      "loss": 0.7925,
+      "step": 136
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999788759893545,
+      "loss": 0.8043,
+      "step": 137
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999785004721968,
+      "loss": 0.7465,
+      "step": 138
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999781216469942,
+      "loss": 0.8322,
+      "step": 139
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997773951375926,
+      "loss": 0.8788,
+      "step": 140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997735407250455,
+      "loss": 0.7898,
+      "step": 141
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997696532324293,
+      "loss": 0.8559,
+      "step": 142
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999765732659872,
+      "loss": 0.7981,
+      "step": 143
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997617790075035,
+      "loss": 0.7904,
+      "step": 144
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997577922754544,
+      "loss": 0.8208,
+      "step": 145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999753772463857,
+      "loss": 0.8139,
+      "step": 146
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999749719572844,
+      "loss": 0.9067,
+      "step": 147
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997456336025496,
+      "loss": 0.8134,
+      "step": 148
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999741514553109,
+      "loss": 0.8543,
+      "step": 149
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997373624246583,
+      "loss": 0.812,
+      "step": 150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019997331772173352,
+      "loss": 0.8334,
+      "step": 151
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997289589312778,
+      "loss": 0.807,
+      "step": 152
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997247075666263,
+      "loss": 0.7596,
+      "step": 153
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997204231235208,
+      "loss": 0.8635,
+      "step": 154
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999716105602103,
+      "loss": 0.7812,
+      "step": 155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999711755002516,
+      "loss": 0.803,
+      "step": 156
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999707371324904,
+      "loss": 0.8068,
+      "step": 157
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997029545694116,
+      "loss": 0.8479,
+      "step": 158
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996985047361854,
+      "loss": 0.8064,
+      "step": 159
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999694021825372,
+      "loss": 0.7946,
+      "step": 160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000199968950583712,
+      "loss": 0.8704,
+      "step": 161
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996849567715792,
+      "loss": 0.8403,
+      "step": 162
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996803746288996,
+      "loss": 0.8898,
+      "step": 163
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996757594092332,
+      "loss": 0.7788,
+      "step": 164
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996711111127326,
+      "loss": 0.84,
+      "step": 165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996664297395515,
+      "loss": 0.7992,
+      "step": 166
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999661715289845,
+      "loss": 0.8083,
+      "step": 167
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996569677637686,
+      "loss": 0.7838,
+      "step": 168
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000199965218716148,
+      "loss": 0.7995,
+      "step": 169
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996473734831366,
+      "loss": 0.8224,
+      "step": 170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996425267288984,
+      "loss": 0.8309,
+      "step": 171
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999637646898926,
+      "loss": 0.8162,
+      "step": 172
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996327339933797,
+      "loss": 0.8214,
+      "step": 173
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996277880124233,
+      "loss": 0.8864,
+      "step": 174
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996228089562196,
+      "loss": 0.8201,
+      "step": 175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996177968249334,
+      "loss": 0.8356,
+      "step": 176
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996127516187312,
+      "loss": 0.7681,
+      "step": 177
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996076733377796,
+      "loss": 0.8029,
+      "step": 178
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999602561982246,
+      "loss": 0.7971,
+      "step": 179
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995974175523003,
+      "loss": 0.813,
+      "step": 180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999592240048113,
+      "loss": 0.8229,
+      "step": 181
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995870294698542,
+      "loss": 0.7887,
+      "step": 182
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995817858176973,
+      "loss": 0.7867,
+      "step": 183
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995765090918157,
+      "loss": 0.7972,
+      "step": 184
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995711992923834,
+      "loss": 0.7363,
+      "step": 185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995658564195767,
+      "loss": 0.7376,
+      "step": 186
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995604804735718,
+      "loss": 0.8753,
+      "step": 187
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995550714545473,
+      "loss": 0.821,
+      "step": 188
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999549629362682,
+      "loss": 0.8063,
+      "step": 189
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995441541981553,
+      "loss": 0.8182,
+      "step": 190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995386459611495,
+      "loss": 0.7914,
+      "step": 191
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995331046518454,
+      "loss": 0.803,
+      "step": 192
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995275302704278,
+      "loss": 0.7851,
+      "step": 193
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995219228170798,
+      "loss": 0.7648,
+      "step": 194
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019995162822919883,
+      "loss": 0.7654,
+      "step": 195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019995106086953388,
+      "loss": 0.8251,
+      "step": 196
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019995049020273196,
+      "loss": 0.8753,
+      "step": 197
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994991622881194,
+      "loss": 0.8021,
+      "step": 198
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999493389477928,
+      "loss": 0.8129,
+      "step": 199
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994875835969368,
+      "loss": 0.8119,
+      "step": 200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994817446453373,
+      "loss": 0.8364,
+      "step": 201
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994758726233234,
+      "loss": 0.8288,
+      "step": 202
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994699675310887,
+      "loss": 0.8751,
+      "step": 203
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999464029368829,
+      "loss": 0.8395,
+      "step": 204
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994580581367407,
+      "loss": 0.7927,
+      "step": 205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994520538350213,
+      "loss": 0.7563,
+      "step": 206
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994460164638692,
+      "loss": 0.7904,
+      "step": 207
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999439946023485,
+      "loss": 0.7722,
+      "step": 208
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994338425140688,
+      "loss": 0.8949,
+      "step": 209
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999427705935823,
+      "loss": 0.7898,
+      "step": 210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994215362889502,
+      "loss": 0.8372,
+      "step": 211
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994153335736546,
+      "loss": 0.8537,
+      "step": 212
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019994090977901418,
+      "loss": 0.8521,
+      "step": 213
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999402828938618,
+      "loss": 0.8168,
+      "step": 214
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993965270192904,
+      "loss": 0.7893,
+      "step": 215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999390192032368,
+      "loss": 0.7778,
+      "step": 216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993838239780595,
+      "loss": 0.8003,
+      "step": 217
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993774228565766,
+      "loss": 0.7848,
+      "step": 218
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993709886681303,
+      "loss": 0.8466,
+      "step": 219
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999364521412934,
+      "loss": 0.8486,
+      "step": 220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993580210912016,
+      "loss": 0.783,
+      "step": 221
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993514877031485,
+      "loss": 0.7472,
+      "step": 222
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993449212489902,
+      "loss": 0.8327,
+      "step": 223
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999338321728944,
+      "loss": 0.8215,
+      "step": 224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999331689143229,
+      "loss": 0.892,
+      "step": 225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993250234920636,
+      "loss": 0.775,
+      "step": 226
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993183247756695,
+      "loss": 0.7259,
+      "step": 227
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019993115929942677,
+      "loss": 0.7449,
+      "step": 228
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999304828148081,
+      "loss": 0.7715,
+      "step": 229
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999298030237333,
+      "loss": 0.8312,
+      "step": 230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999291199262249,
+      "loss": 0.8275,
+      "step": 231
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019992843352230547,
+      "loss": 0.7595,
+      "step": 232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019992774381199778,
+      "loss": 0.9094,
+      "step": 233
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999270507953246,
+      "loss": 0.8676,
+      "step": 234
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019992635447230886,
+      "loss": 0.7904,
+      "step": 235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019992565484297362,
+      "loss": 0.8133,
+      "step": 236
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019992495190734202,
+      "loss": 0.812,
+      "step": 237
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019992424566543733,
+      "loss": 0.8083,
+      "step": 238
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019992353611728288,
+      "loss": 0.8047,
+      "step": 239
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999228232629022,
+      "loss": 0.8367,
+      "step": 240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019992210710231885,
+      "loss": 0.8631,
+      "step": 241
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019992138763555653,
+      "loss": 0.8129,
+      "step": 242
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199920664862639,
+      "loss": 0.8135,
+      "step": 243
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991993878359026,
+      "loss": 0.7993,
+      "step": 244
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991920939843428,
+      "loss": 0.8342,
+      "step": 245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991847670719522,
+      "loss": 0.8817,
+      "step": 246
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991774070989728,
+      "loss": 0.7526,
+      "step": 247
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991700140656486,
+      "loss": 0.7974,
+      "step": 248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999162587972224,
+      "loss": 0.7931,
+      "step": 249
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991551288189448,
+      "loss": 0.7768,
+      "step": 250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991476366060576,
+      "loss": 0.7896,
+      "step": 251
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991401113338104,
+      "loss": 0.8606,
+      "step": 252
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991325530024524,
+      "loss": 0.7774,
+      "step": 253
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991249616122333,
+      "loss": 0.8041,
+      "step": 254
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999117337163405,
+      "loss": 0.7721,
+      "step": 255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991096796562188,
+      "loss": 0.9457,
+      "step": 256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991019890909284,
+      "loss": 0.7721,
+      "step": 257
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990942654677887,
+      "loss": 0.7819,
+      "step": 258
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999086508787055,
+      "loss": 0.8086,
+      "step": 259
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999078719048984,
+      "loss": 0.7654,
+      "step": 260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999070896253833,
+      "loss": 0.8203,
+      "step": 261
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990630404018614,
+      "loss": 0.8168,
+      "step": 262
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999055151493329,
+      "loss": 0.8178,
+      "step": 263
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990472295284962,
+      "loss": 0.7597,
+      "step": 264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990392745076262,
+      "loss": 0.7857,
+      "step": 265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990312864309818,
+      "loss": 0.8042,
+      "step": 266
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990232652988268,
+      "loss": 0.8358,
+      "step": 267
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999015211111427,
+      "loss": 0.7333,
+      "step": 268
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999007123869049,
+      "loss": 0.8118,
+      "step": 269
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989990035719602,
+      "loss": 0.7939,
+      "step": 270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989908502204292,
+      "loss": 0.8264,
+      "step": 271
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989826638147263,
+      "loss": 0.8698,
+      "step": 272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989744443551217,
+      "loss": 0.7266,
+      "step": 273
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001998966191841888,
+      "loss": 0.768,
+      "step": 274
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989579062752974,
+      "loss": 0.8611,
+      "step": 275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001998949587655625,
+      "loss": 0.808,
+      "step": 276
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989412359831455,
+      "loss": 0.7849,
+      "step": 277
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001998932851258135,
+      "loss": 0.8627,
+      "step": 278
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989244334808715,
+      "loss": 0.7497,
+      "step": 279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989159826516332,
+      "loss": 0.6951,
+      "step": 280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019989074987706998,
+      "loss": 0.756,
+      "step": 281
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988989818383522,
+      "loss": 0.7924,
+      "step": 282
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998890431854872,
+      "loss": 0.8103,
+      "step": 283
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988818488205417,
+      "loss": 0.7668,
+      "step": 284
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988732327356462,
+      "loss": 0.7947,
+      "step": 285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199886458360047,
+      "loss": 0.7875,
+      "step": 286
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988559014152992,
+      "loss": 0.8308,
+      "step": 287
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988471861804213,
+      "loss": 0.8544,
+      "step": 288
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988384378961246,
+      "loss": 0.7925,
+      "step": 289
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988296565626987,
+      "loss": 0.7661,
+      "step": 290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988208421804337,
+      "loss": 0.9,
+      "step": 291
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998811994749622,
+      "loss": 0.7279,
+      "step": 292
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988031142705558,
+      "loss": 0.7891,
+      "step": 293
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987942007435288,
+      "loss": 0.7736,
+      "step": 294
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987852541688363,
+      "loss": 0.8087,
+      "step": 295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998776274546774,
+      "loss": 0.7975,
+      "step": 296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987672618776396,
+      "loss": 0.8358,
+      "step": 297
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998758216161731,
+      "loss": 0.8448,
+      "step": 298
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987491373993469,
+      "loss": 0.7933,
+      "step": 299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987400255907885,
+      "loss": 0.7753,
+      "step": 300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998730880736357,
+      "loss": 0.7666,
+      "step": 301
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987217028363547,
+      "loss": 0.8375,
+      "step": 302
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987124918910857,
+      "loss": 0.8066,
+      "step": 303
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987032479008547,
+      "loss": 0.7497,
+      "step": 304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986939708659677,
+      "loss": 0.8451,
+      "step": 305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998684660786731,
+      "loss": 0.7709,
+      "step": 306
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998675317663453,
+      "loss": 0.814,
+      "step": 307
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986659414964432,
+      "loss": 0.8008,
+      "step": 308
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986565322860115,
+      "loss": 0.7649,
+      "step": 309
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986470900324694,
+      "loss": 0.8151,
+      "step": 310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986376147361292,
+      "loss": 0.8009,
+      "step": 311
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986281063973042,
+      "loss": 0.8264,
+      "step": 312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986185650163095,
+      "loss": 0.7429,
+      "step": 313
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986089905934601,
+      "loss": 0.7226,
+      "step": 314
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985993831290734,
+      "loss": 0.7465,
+      "step": 315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985897426234674,
+      "loss": 0.8774,
+      "step": 316
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985800690769606,
+      "loss": 0.8039,
+      "step": 317
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985703624898733,
+      "loss": 0.7816,
+      "step": 318
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985606228625264,
+      "loss": 0.7705,
+      "step": 319
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985508501952425,
+      "loss": 0.7647,
+      "step": 320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998541044488345,
+      "loss": 0.869,
+      "step": 321
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985312057421584,
+      "loss": 0.7976,
+      "step": 322
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985213339570077,
+      "loss": 0.7911,
+      "step": 323
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019985114291332199,
+      "loss": 0.7574,
+      "step": 324
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019985014912711228,
+      "loss": 0.7837,
+      "step": 325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998491520371045,
+      "loss": 0.8448,
+      "step": 326
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019984815164333163,
+      "loss": 0.8004,
+      "step": 327
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019984714794582683,
+      "loss": 0.7659,
+      "step": 328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019984614094462324,
+      "loss": 0.8168,
+      "step": 329
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019984513063975423,
+      "loss": 0.7989,
+      "step": 330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019984411703125322,
+      "loss": 0.8476,
+      "step": 331
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019984310011915372,
+      "loss": 0.8231,
+      "step": 332
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998420799034894,
+      "loss": 0.7966,
+      "step": 333
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019984105638429402,
+      "loss": 0.8221,
+      "step": 334
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019984002956160147,
+      "loss": 0.7966,
+      "step": 335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019983899943544565,
+      "loss": 0.8054,
+      "step": 336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998379660058607,
+      "loss": 0.9002,
+      "step": 337
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998369292728808,
+      "loss": 0.8068,
+      "step": 338
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019983588923654024,
+      "loss": 0.8665,
+      "step": 339
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019983484589687346,
+      "loss": 0.8747,
+      "step": 340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019983379925391493,
+      "loss": 0.8186,
+      "step": 341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019983274930769938,
+      "loss": 0.7857,
+      "step": 342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019983169605826146,
+      "loss": 0.8397,
+      "step": 343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019983063950563604,
+      "loss": 0.794,
+      "step": 344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019982957964985812,
+      "loss": 0.7846,
+      "step": 345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998285164909627,
+      "loss": 0.7461,
+      "step": 346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000199827450028985,
+      "loss": 0.7788,
+      "step": 347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019982638026396033,
+      "loss": 0.8395,
+      "step": 348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019982530719592402,
+      "loss": 0.7953,
+      "step": 349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019982423082491162,
+      "loss": 0.747,
+      "step": 350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019982315115095873,
+      "loss": 0.8333,
+      "step": 351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019982206817410107,
+      "loss": 0.8411,
+      "step": 352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998209818943745,
+      "loss": 0.8851,
+      "step": 353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019981989231181494,
+      "loss": 0.7965,
+      "step": 354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019981879942645845,
+      "loss": 0.7353,
+      "step": 355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019981770323834118,
+      "loss": 0.848,
+      "step": 356
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998166037474994,
+      "loss": 0.7833,
+      "step": 357
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998155009539695,
+      "loss": 0.7652,
+      "step": 358
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019981439485778797,
+      "loss": 0.7925,
+      "step": 359
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998132854589914,
+      "loss": 0.7999,
+      "step": 360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001998121727576165,
+      "loss": 0.7903,
+      "step": 361
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019981105675370006,
+      "loss": 0.764,
+      "step": 362
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019980993744727908,
+      "loss": 0.8055,
+      "step": 363
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019980881483839052,
+      "loss": 0.7711,
+      "step": 364
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019980768892707157,
+      "loss": 0.805,
+      "step": 365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019980655971335945,
+      "loss": 0.8102,
+      "step": 366
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019980542719729154,
+      "loss": 0.7908,
+      "step": 367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001998042913789053,
+      "loss": 0.7892,
+      "step": 368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019980315225823832,
+      "loss": 0.8428,
+      "step": 369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019980200983532833,
+      "loss": 0.8324,
+      "step": 370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019980086411021305,
+      "loss": 0.7979,
+      "step": 371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019979971508293047,
+      "loss": 0.7712,
+      "step": 372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019979856275351856,
+      "loss": 0.7714,
+      "step": 373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019979740712201545,
+      "loss": 0.857,
+      "step": 374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001997962481884594,
+      "loss": 0.8727,
+      "step": 375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019979508595288872,
+      "loss": 0.7743,
+      "step": 376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019979392041534187,
+      "loss": 0.8434,
+      "step": 377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019979275157585747,
+      "loss": 0.8585,
+      "step": 378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019979157943447414,
+      "loss": 0.8423,
+      "step": 379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001997904039912307,
+      "loss": 0.7517,
+      "step": 380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000199789225246166,
+      "loss": 0.8478,
+      "step": 381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001997880431993191,
+      "loss": 0.8557,
+      "step": 382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019978685785072904,
+      "loss": 0.7752,
+      "step": 383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001997856692004351,
+      "loss": 0.8654,
+      "step": 384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019978447724847652,
+      "loss": 0.8535,
+      "step": 385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019978328199489286,
+      "loss": 0.7456,
+      "step": 386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019978208343972363,
+      "loss": 0.7465,
+      "step": 387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001997808815830084,
+      "loss": 0.7754,
+      "step": 388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019977967642478706,
+      "loss": 0.8185,
+      "step": 389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019977846796509943,
+      "loss": 0.7896,
+      "step": 390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019977725620398546,
+      "loss": 0.737,
+      "step": 391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019977604114148533,
+      "loss": 0.7979,
+      "step": 392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019977482277763918,
+      "loss": 0.8144,
+      "step": 393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019977360111248727,
+      "loss": 0.8341,
+      "step": 394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019977237614607018,
+      "loss": 0.8033,
+      "step": 395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001997711478784283,
+      "loss": 0.7847,
+      "step": 396
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019976991630960234,
+      "loss": 0.7661,
+      "step": 397
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000199768681439633,
+      "loss": 0.7665,
+      "step": 398
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019976744326856118,
+      "loss": 0.7943,
+      "step": 399
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019976620179642786,
+      "loss": 0.7185,
+      "step": 400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019976495702327406,
+      "loss": 0.7544,
+      "step": 401
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019976370894914106,
+      "loss": 0.7819,
+      "step": 402
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019976245757407003,
+      "loss": 0.8308,
+      "step": 403
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019976120289810247,
+      "loss": 0.8323,
+      "step": 404
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019975994492127988,
+      "loss": 0.7911,
+      "step": 405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019975868364364383,
+      "loss": 0.7838,
+      "step": 406
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001997574190652361,
+      "loss": 0.7548,
+      "step": 407
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019975615118609854,
+      "loss": 0.8319,
+      "step": 408
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001997548800062731,
+      "loss": 0.8818,
+      "step": 409
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019975360552580178,
+      "loss": 0.8119,
+      "step": 410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975232774472682,
+      "loss": 0.8118,
+      "step": 411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019975104666309048,
+      "loss": 0.8313,
+      "step": 412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997497622809351,
+      "loss": 0.7909,
+      "step": 413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974847459830326,
+      "loss": 0.8009,
+      "step": 414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997471836152375,
+      "loss": 0.8188,
+      "step": 415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974588933178056,
+      "loss": 0.8074,
+      "step": 416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997445917479753,
+      "loss": 0.7901,
+      "step": 417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974329086386457,
+      "loss": 0.7777,
+      "step": 418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997419866794915,
+      "loss": 0.7863,
+      "step": 419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019974067919489917,
+      "loss": 0.8648,
+      "step": 420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973936841013087,
+      "loss": 0.7781,
+      "step": 421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973805432522997,
+      "loss": 0.7974,
+      "step": 422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973673694024,
+      "loss": 0.825,
+      "step": 423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997354162552045,
+      "loss": 0.7407,
+      "step": 424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973409227016714,
+      "loss": 0.7754,
+      "step": 425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973276498517177,
+      "loss": 0.7507,
+      "step": 426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997314344002623,
+      "loss": 0.7631,
+      "step": 427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019973010051548275,
+      "loss": 0.8441,
+      "step": 428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019972876333087728,
+      "loss": 0.8574,
+      "step": 429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019972742284649008,
+      "loss": 0.8064,
+      "step": 430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019972607906236555,
+      "loss": 0.8711,
+      "step": 431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019972473197854811,
+      "loss": 0.8583,
+      "step": 432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019972338159508237,
+      "loss": 0.7646,
+      "step": 433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019972202791201303,
+      "loss": 0.7415,
+      "step": 434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997206709293848,
+      "loss": 0.8165,
+      "step": 435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997193106472427,
+      "loss": 0.7906,
+      "step": 436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997179470656316,
+      "loss": 0.7006,
+      "step": 437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997165801845967,
+      "loss": 0.7607,
+      "step": 438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019971521000418323,
+      "loss": 0.7851,
+      "step": 439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997138365244365,
+      "loss": 0.8041,
+      "step": 440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019971245974540196,
+      "loss": 0.8188,
+      "step": 441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019971107966712518,
+      "loss": 0.8126,
+      "step": 442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019970969628965179,
+      "loss": 0.7739,
+      "step": 443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001997083096130276,
+      "loss": 0.8247,
+      "step": 444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019970691963729846,
+      "loss": 0.851,
+      "step": 445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019970552636251038,
+      "loss": 0.7674,
+      "step": 446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019970412978870947,
+      "loss": 0.8429,
+      "step": 447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019970272991594193,
+      "loss": 0.763,
+      "step": 448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019970132674425407,
+      "loss": 0.8225,
+      "step": 449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019969992027369228,
+      "loss": 0.7699,
+      "step": 450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019969851050430316,
+      "loss": 0.7724,
+      "step": 451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019969709743613336,
+      "loss": 0.8066,
+      "step": 452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001996956810692296,
+      "loss": 0.8156,
+      "step": 453
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969426140363874,
+      "loss": 0.705,
+      "step": 454
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969283843940776,
+      "loss": 0.8421,
+      "step": 455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019969141217658377,
+      "loss": 0.7843,
+      "step": 456
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968998261521393,
+      "loss": 0.8125,
+      "step": 457
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968854975534553,
+      "loss": 0.848,
+      "step": 458
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000199687113597026,
+      "loss": 0.8061,
+      "step": 459
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996856741403029,
+      "loss": 0.771,
+      "step": 460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996842313852238,
+      "loss": 0.8071,
+      "step": 461
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019968278533183647,
+      "loss": 0.7877,
+      "step": 462
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996813359801887,
+      "loss": 0.7999,
+      "step": 463
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967988333032853,
+      "loss": 0.8959,
+      "step": 464
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967842738230393,
+      "loss": 0.8302,
+      "step": 465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967696813616317,
+      "loss": 0.7108,
+      "step": 466
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967550559195444,
+      "loss": 0.7529,
+      "step": 467
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967403974972622,
+      "loss": 0.8186,
+      "step": 468
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019967257060952696,
+      "loss": 0.7813,
+      "step": 469
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996710981714053,
+      "loss": 0.7698,
+      "step": 470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019966962243540994,
+      "loss": 0.808,
+      "step": 471
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996681434015897,
+      "loss": 0.8951,
+      "step": 472
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996666610699935,
+      "loss": 0.8085,
+      "step": 473
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019966517544067045,
+      "loss": 0.7963,
+      "step": 474
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019966368651366966,
+      "loss": 0.8229,
+      "step": 475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996621942890404,
+      "loss": 0.7495,
+      "step": 476
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019966069876683204,
+      "loss": 0.7546,
+      "step": 477
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965919994709407,
+      "loss": 0.7445,
+      "step": 478
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996576978298761,
+      "loss": 0.8094,
+      "step": 479
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996561924152278,
+      "loss": 0.7465,
+      "step": 480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965468370319905,
+      "loss": 0.7525,
+      "step": 481
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965317169383964,
+      "loss": 0.7284,
+      "step": 482
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996516563871997,
+      "loss": 0.9095,
+      "step": 483
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019965013778332934,
+      "loss": 0.7865,
+      "step": 484
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964861588227882,
+      "loss": 0.7769,
+      "step": 485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996470906840985,
+      "loss": 0.7217,
+      "step": 486
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996455621888388,
+      "loss": 0.8498,
+      "step": 487
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964403039655036,
+      "loss": 0.7553,
+      "step": 488
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964249530728378,
+      "loss": 0.8226,
+      "step": 489
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019964095692108993,
+      "loss": 0.7887,
+      "step": 490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001996394152380197,
+      "loss": 0.7965,
+      "step": 491
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963787025812404,
+      "loss": 0.8101,
+      "step": 492
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963632198145414,
+      "loss": 0.8037,
+      "step": 493
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963477040806118,
+      "loss": 0.7657,
+      "step": 494
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963321553799653,
+      "loss": 0.8367,
+      "step": 495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019963165737131163,
+      "loss": 0.7941,
+      "step": 496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000199630095908058,
+      "loss": 0.7336,
+      "step": 497
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019962853114828736,
+      "loss": 0.8293,
+      "step": 498
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019962696309205148,
+      "loss": 0.7968,
+      "step": 499
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001996253917394022,
+      "loss": 0.7434,
+      "step": 500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019962381709039153,
+      "loss": 0.8363,
+      "step": 501
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019962223914507158,
+      "loss": 0.7887,
+      "step": 502
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019962065790349458,
+      "loss": 0.8115,
+      "step": 503
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019961907336571278,
+      "loss": 0.7277,
+      "step": 504
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019961748553177868,
+      "loss": 0.78,
+      "step": 505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019961589440174482,
+      "loss": 0.7703,
+      "step": 506
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019961429997566378,
+      "loss": 0.741,
+      "step": 507
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019961270225358837,
+      "loss": 0.7245,
+      "step": 508
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019961110123557142,
+      "loss": 0.836,
+      "step": 509
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019960949692166596,
+      "loss": 0.7567,
+      "step": 510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000199607889311925,
+      "loss": 0.7788,
+      "step": 511
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019960627840640181,
+      "loss": 0.827,
+      "step": 512
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001996046642051496,
+      "loss": 0.7658,
+      "step": 513
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019960304670822186,
+      "loss": 0.7505,
+      "step": 514
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019960142591567208,
+      "loss": 0.8378,
+      "step": 515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019959980182755385,
+      "loss": 0.8108,
+      "step": 516
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000199598174443921,
+      "loss": 0.7649,
+      "step": 517
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995965437648273,
+      "loss": 0.753,
+      "step": 518
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995949097903267,
+      "loss": 0.8634,
+      "step": 519
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019959327252047328,
+      "loss": 0.7534,
+      "step": 520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019959163195532124,
+      "loss": 0.8162,
+      "step": 521
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958998809492485,
+      "loss": 0.7938,
+      "step": 522
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995883409393385,
+      "loss": 0.8145,
+      "step": 523
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995866904886167,
+      "loss": 0.7788,
+      "step": 524
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958503674281399,
+      "loss": 0.745,
+      "step": 525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995833797019852,
+      "loss": 0.7502,
+      "step": 526
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958171936618508,
+      "loss": 0.7794,
+      "step": 527
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019958005573546855,
+      "loss": 0.7664,
+      "step": 528
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957838880989078,
+      "loss": 0.7641,
+      "step": 529
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957671858950677,
+      "loss": 0.8056,
+      "step": 530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957504507437185,
+      "loss": 0.7491,
+      "step": 531
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957336826454143,
+      "loss": 0.862,
+      "step": 532
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957168816007092,
+      "loss": 0.8399,
+      "step": 533
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019957000476101598,
+      "loss": 0.7734,
+      "step": 534
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956831806743227,
+      "loss": 0.7558,
+      "step": 535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956662807937557,
+      "loss": 0.7223,
+      "step": 536
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001995649347969019,
+      "loss": 0.7795,
+      "step": 537
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956323822006714,
+      "loss": 0.8093,
+      "step": 538
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019956153834892754,
+      "loss": 0.8192,
+      "step": 539
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001995598351835393,
+      "loss": 0.7119,
+      "step": 540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019955812872395878,
+      "loss": 0.723,
+      "step": 541
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019955641897024246,
+      "loss": 0.8054,
+      "step": 542
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019955470592244686,
+      "loss": 0.8095,
+      "step": 543
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019955298958062876,
+      "loss": 0.8501,
+      "step": 544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019955126994484484,
+      "loss": 0.7943,
+      "step": 545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019954954701515204,
+      "loss": 0.7487,
+      "step": 546
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001995478207916074,
+      "loss": 0.7855,
+      "step": 547
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000199546091274268,
+      "loss": 0.7582,
+      "step": 548
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019954435846319108,
+      "loss": 0.8002,
+      "step": 549
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019954262235843395,
+      "loss": 0.754,
+      "step": 550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001995408829600541,
+      "loss": 0.7934,
+      "step": 551
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019953914026810902,
+      "loss": 0.7639,
+      "step": 552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019953739428265642,
+      "loss": 0.7333,
+      "step": 553
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019953564500375408,
+      "loss": 0.754,
+      "step": 554
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019953389243145982,
+      "loss": 0.8023,
+      "step": 555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019953213656583168,
+      "loss": 0.8217,
+      "step": 556
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019953037740692774,
+      "loss": 0.7119,
+      "step": 557
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001995286149548062,
+      "loss": 0.8151,
+      "step": 558
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001995268492095254,
+      "loss": 0.6971,
+      "step": 559
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019952508017114374,
+      "loss": 0.7958,
+      "step": 560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019952330783971976,
+      "loss": 0.7203,
+      "step": 561
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001995215322153121,
+      "loss": 0.7438,
+      "step": 562
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019951975329797952,
+      "loss": 0.7445,
+      "step": 563
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019951797108778087,
+      "loss": 0.8638,
+      "step": 564
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019951618558477515,
+      "loss": 0.829,
+      "step": 565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019951439678902135,
+      "loss": 0.8351,
+      "step": 566
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019951260470057878,
+      "loss": 0.747,
+      "step": 567
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019951080931950664,
+      "loss": 0.766,
+      "step": 568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019950901064586435,
+      "loss": 0.8215,
+      "step": 569
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019950720867971148,
+      "loss": 0.7948,
+      "step": 570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001995054034211076,
+      "loss": 0.8025,
+      "step": 571
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019950359487011244,
+      "loss": 0.8184,
+      "step": 572
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019950359487011244,
+      "loss": 0.817,
+      "step": 573
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001995017830267859,
+      "loss": 0.7725,
+      "step": 574
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019949996789118786,
+      "loss": 0.7854,
+      "step": 575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019949814946337838,
+      "loss": 0.8401,
+      "step": 576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019949632774341768,
+      "loss": 0.8023,
+      "step": 577
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019949450273136602,
+      "loss": 0.8514,
+      "step": 578
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019949267442728378,
+      "loss": 0.7743,
+      "step": 579
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001994908428312314,
+      "loss": 0.7445,
+      "step": 580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019948900794326956,
+      "loss": 0.7645,
+      "step": 581
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019948716976345897,
+      "loss": 0.7685,
+      "step": 582
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019948532829186036,
+      "loss": 0.8071,
+      "step": 583
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019948348352853476,
+      "loss": 0.7766,
+      "step": 584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019948163547354317,
+      "loss": 0.7322,
+      "step": 585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019947978412694674,
+      "loss": 0.8139,
+      "step": 586
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019947792948880674,
+      "loss": 0.7984,
+      "step": 587
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001994760715591845,
+      "loss": 0.7774,
+      "step": 588
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001994742103381415,
+      "loss": 0.8295,
+      "step": 589
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019947234582573937,
+      "loss": 0.7993,
+      "step": 590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019947047802203977,
+      "loss": 0.8016,
+      "step": 591
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001994686069271045,
+      "loss": 0.7452,
+      "step": 592
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019946673254099545,
+      "loss": 0.7798,
+      "step": 593
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001994648548637747,
+      "loss": 0.6703,
+      "step": 594
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019946297389550433,
+      "loss": 0.7192,
+      "step": 595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019946108963624654,
+      "loss": 0.7675,
+      "step": 596
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019945920208606376,
+      "loss": 0.8103,
+      "step": 597
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019945731124501842,
+      "loss": 0.7503,
+      "step": 598
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019945541711317307,
+      "loss": 0.8073,
+      "step": 599
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019945351969059033,
+      "loss": 0.8421,
+      "step": 600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019945161897733308,
+      "loss": 0.7399,
+      "step": 601
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019944971497346415,
+      "loss": 0.7475,
+      "step": 602
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019944780767904653,
+      "loss": 0.7655,
+      "step": 603
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019944589709414338,
+      "loss": 0.7986,
+      "step": 604
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019944398321881788,
+      "loss": 0.7682,
+      "step": 605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019944206605313335,
+      "loss": 0.7835,
+      "step": 606
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019944014559715326,
+      "loss": 0.7748,
+      "step": 607
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001994382218509411,
+      "loss": 0.8386,
+      "step": 608
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019943629481456053,
+      "loss": 0.817,
+      "step": 609
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019943436448807537,
+      "loss": 0.7912,
+      "step": 610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019943243087154944,
+      "loss": 0.814,
+      "step": 611
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019943049396504674,
+      "loss": 0.7438,
+      "step": 612
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019942855376863135,
+      "loss": 0.791,
+      "step": 613
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019942661028236745,
+      "loss": 0.7691,
+      "step": 614
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019942466350631935,
+      "loss": 0.7843,
+      "step": 615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019942271344055147,
+      "loss": 0.8026,
+      "step": 616
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019942076008512837,
+      "loss": 0.7579,
+      "step": 617
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019941880344011463,
+      "loss": 0.7768,
+      "step": 618
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019941684350557502,
+      "loss": 0.8056,
+      "step": 619
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019941488028157438,
+      "loss": 0.8091,
+      "step": 620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019941291376817763,
+      "loss": 0.7535,
+      "step": 621
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001994109439654499,
+      "loss": 0.7866,
+      "step": 622
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019940897087345632,
+      "loss": 0.8244,
+      "step": 623
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019940699449226223,
+      "loss": 0.7662,
+      "step": 624
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019940501482193298,
+      "loss": 0.7547,
+      "step": 625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019940303186253403,
+      "loss": 0.8408,
+      "step": 626
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001994010456141311,
+      "loss": 0.8075,
+      "step": 627
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019939905607678984,
+      "loss": 0.8162,
+      "step": 628
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019939706325057608,
+      "loss": 0.756,
+      "step": 629
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993950671355558,
+      "loss": 0.8613,
+      "step": 630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019939306773179497,
+      "loss": 0.8017,
+      "step": 631
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993910650393598,
+      "loss": 0.8012,
+      "step": 632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019938905905831654,
+      "loss": 0.8359,
+      "step": 633
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019938704978873158,
+      "loss": 0.7421,
+      "step": 634
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993850372306714,
+      "loss": 0.8326,
+      "step": 635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019938302138420254,
+      "loss": 0.779,
+      "step": 636
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019938100224939177,
+      "loss": 0.7669,
+      "step": 637
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993789798263059,
+      "loss": 0.7357,
+      "step": 638
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019937695411501177,
+      "loss": 0.7885,
+      "step": 639
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019937492511557645,
+      "loss": 0.7785,
+      "step": 640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993728928280671,
+      "loss": 0.763,
+      "step": 641
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993708572525509,
+      "loss": 0.7752,
+      "step": 642
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993688183890953,
+      "loss": 0.8238,
+      "step": 643
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019936677623776764,
+      "loss": 0.7393,
+      "step": 644
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019936473079863558,
+      "loss": 0.8016,
+      "step": 645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993626820717668,
+      "loss": 0.7891,
+      "step": 646
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000199360630057229,
+      "loss": 0.7593,
+      "step": 647
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019935857475509016,
+      "loss": 0.7439,
+      "step": 648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019935651616541828,
+      "loss": 0.8218,
+      "step": 649
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019935445428828142,
+      "loss": 0.7491,
+      "step": 650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019935238912374787,
+      "loss": 0.7488,
+      "step": 651
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993503206718859,
+      "loss": 0.7972,
+      "step": 652
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019934824893276399,
+      "loss": 0.8319,
+      "step": 653
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019934617390645064,
+      "loss": 0.791,
+      "step": 654
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019934409559301458,
+      "loss": 0.6857,
+      "step": 655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019934201399252454,
+      "loss": 0.8472,
+      "step": 656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019933992910504936,
+      "loss": 0.8374,
+      "step": 657
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993378409306581,
+      "loss": 0.8097,
+      "step": 658
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993357494694198,
+      "loss": 0.793,
+      "step": 659
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019933365472140366,
+      "loss": 0.7916,
+      "step": 660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000199331556686679,
+      "loss": 0.8299,
+      "step": 661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019932945536531524,
+      "loss": 0.7425,
+      "step": 662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019932735075738195,
+      "loss": 0.8046,
+      "step": 663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019932524286294867,
+      "loss": 0.7312,
+      "step": 664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019932313168208524,
+      "loss": 0.8265,
+      "step": 665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019932101721486145,
+      "loss": 0.8185,
+      "step": 666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993188994613473,
+      "loss": 0.7705,
+      "step": 667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001993167784216128,
+      "loss": 0.7967,
+      "step": 668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019931465409572826,
+      "loss": 0.7607,
+      "step": 669
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019931252648376383,
+      "loss": 0.7873,
+      "step": 670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019931039558578997,
+      "loss": 0.7908,
+      "step": 671
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019930826140187718,
+      "loss": 0.8388,
+      "step": 672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001993061239320961,
+      "loss": 0.7623,
+      "step": 673
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019930398317651743,
+      "loss": 0.6908,
+      "step": 674
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019930183913521195,
+      "loss": 0.8133,
+      "step": 675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019929969180825066,
+      "loss": 0.7362,
+      "step": 676
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019929754119570465,
+      "loss": 0.7381,
+      "step": 677
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000199295387297645,
+      "loss": 0.8052,
+      "step": 678
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019929323011414304,
+      "loss": 0.8315,
+      "step": 679
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019929106964527006,
+      "loss": 0.7258,
+      "step": 680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019928890589109764,
+      "loss": 0.8093,
+      "step": 681
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019928673885169734,
+      "loss": 0.7842,
+      "step": 682
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019928456852714082,
+      "loss": 0.805,
+      "step": 683
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019928239491749994,
+      "loss": 0.7272,
+      "step": 684
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001992802180228466,
+      "loss": 0.8247,
+      "step": 685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019927803784325284,
+      "loss": 0.7424,
+      "step": 686
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019927585437879078,
+      "loss": 0.7458,
+      "step": 687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001992736676295327,
+      "loss": 0.7504,
+      "step": 688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019927147759555089,
+      "loss": 0.8167,
+      "step": 689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019926928427691786,
+      "loss": 0.7904,
+      "step": 690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001992670876737062,
+      "loss": 0.7182,
+      "step": 691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019926488778598852,
+      "loss": 0.8029,
+      "step": 692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019926268461383767,
+      "loss": 0.803,
+      "step": 693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019926047815732655,
+      "loss": 0.7787,
+      "step": 694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019925826841652816,
+      "loss": 0.8045,
+      "step": 695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019925605539151557,
+      "loss": 0.7678,
+      "step": 696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019925383908236203,
+      "loss": 0.8094,
+      "step": 697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019925161948914087,
+      "loss": 0.7966,
+      "step": 698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001992493966119256,
+      "loss": 0.8275,
+      "step": 699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019924717045078963,
+      "loss": 0.8449,
+      "step": 700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019924494100580673,
+      "loss": 0.7239,
+      "step": 701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019924270827705062,
+      "loss": 0.7668,
+      "step": 702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001992404722645952,
+      "loss": 0.8301,
+      "step": 703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019923823296851448,
+      "loss": 0.8099,
+      "step": 704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019923599038888245,
+      "loss": 0.7131,
+      "step": 705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001992337445257734,
+      "loss": 0.7914,
+      "step": 706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019923149537926164,
+      "loss": 0.7744,
+      "step": 707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019922924294942153,
+      "loss": 0.8006,
+      "step": 708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019922698723632767,
+      "loss": 0.7865,
+      "step": 709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019922472824005462,
+      "loss": 0.7767,
+      "step": 710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001992224659606772,
+      "loss": 0.8612,
+      "step": 711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001992202003982702,
+      "loss": 0.8042,
+      "step": 712
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001992179315529086,
+      "loss": 0.7669,
+      "step": 713
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019921565942466752,
+      "loss": 0.8057,
+      "step": 714
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001992133840136221,
+      "loss": 0.8556,
+      "step": 715
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001992111053198476,
+      "loss": 0.7741,
+      "step": 716
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019920882334341944,
+      "loss": 0.7159,
+      "step": 717
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019920653808441313,
+      "loss": 0.78,
+      "step": 718
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001992042495429043,
+      "loss": 0.7867,
+      "step": 719
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019920195771896864,
+      "loss": 0.7997,
+      "step": 720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019919966261268198,
+      "loss": 0.7343,
+      "step": 721
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991973642241203,
+      "loss": 0.8196,
+      "step": 722
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019919506255335962,
+      "loss": 0.8141,
+      "step": 723
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991927576004761,
+      "loss": 0.7357,
+      "step": 724
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000199190449365546,
+      "loss": 0.8056,
+      "step": 725
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019918813784864568,
+      "loss": 0.8003,
+      "step": 726
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019918582304985166,
+      "loss": 0.7424,
+      "step": 727
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991835049692405,
+      "loss": 0.8036,
+      "step": 728
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019918118360688893,
+      "loss": 0.8286,
+      "step": 729
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019917885896287373,
+      "loss": 0.7648,
+      "step": 730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019917653103727182,
+      "loss": 0.7906,
+      "step": 731
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019917419983016025,
+      "loss": 0.7334,
+      "step": 732
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019917186534161612,
+      "loss": 0.7937,
+      "step": 733
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991695275717167,
+      "loss": 0.8081,
+      "step": 734
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991671865205393,
+      "loss": 0.7708,
+      "step": 735
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019916484218816143,
+      "loss": 0.7813,
+      "step": 736
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019916249457466065,
+      "loss": 0.799,
+      "step": 737
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991601436801146,
+      "loss": 0.7659,
+      "step": 738
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019915778950460112,
+      "loss": 0.7994,
+      "step": 739
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019915543204819804,
+      "loss": 0.8363,
+      "step": 740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991530713109834,
+      "loss": 0.7872,
+      "step": 741
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019915070729303533,
+      "loss": 0.7848,
+      "step": 742
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000199148339994432,
+      "loss": 0.775,
+      "step": 743
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991459694152518,
+      "loss": 0.8304,
+      "step": 744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991435955555731,
+      "loss": 0.7455,
+      "step": 745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991412184154745,
+      "loss": 0.7586,
+      "step": 746
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991388379950346,
+      "loss": 0.7687,
+      "step": 747
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019913645429433223,
+      "loss": 0.7846,
+      "step": 748
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019913406731344622,
+      "loss": 0.7956,
+      "step": 749
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991316770524555,
+      "loss": 0.8504,
+      "step": 750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019912928351143927,
+      "loss": 0.8345,
+      "step": 751
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019912688669047665,
+      "loss": 0.7403,
+      "step": 752
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019912448658964695,
+      "loss": 0.7743,
+      "step": 753
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001991220832090296,
+      "loss": 0.7958,
+      "step": 754
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019911967654870415,
+      "loss": 0.73,
+      "step": 755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019911726660875014,
+      "loss": 0.7627,
+      "step": 756
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001991148533892474,
+      "loss": 0.786,
+      "step": 757
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019911243689027573,
+      "loss": 0.7189,
+      "step": 758
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019911001711191514,
+      "loss": 0.8654,
+      "step": 759
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019910759405424562,
+      "loss": 0.8218,
+      "step": 760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001991051677173474,
+      "loss": 0.8006,
+      "step": 761
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019910273810130072,
+      "loss": 0.8133,
+      "step": 762
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000199100305206186,
+      "loss": 0.7501,
+      "step": 763
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019909786903208373,
+      "loss": 0.7765,
+      "step": 764
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990954295790745,
+      "loss": 0.7499,
+      "step": 765
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019909298684723904,
+      "loss": 0.7474,
+      "step": 766
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019909054083665822,
+      "loss": 0.7653,
+      "step": 767
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019908809154741287,
+      "loss": 0.7546,
+      "step": 768
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990856389795841,
+      "loss": 0.848,
+      "step": 769
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990831831332531,
+      "loss": 0.7291,
+      "step": 770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019908072400850104,
+      "loss": 0.8142,
+      "step": 771
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990782616054093,
+      "loss": 0.7355,
+      "step": 772
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990757959240594,
+      "loss": 0.7681,
+      "step": 773
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019907332696453295,
+      "loss": 0.8057,
+      "step": 774
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990708547269115,
+      "loss": 0.751,
+      "step": 775
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000199068379211277,
+      "loss": 0.858,
+      "step": 776
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990659004177113,
+      "loss": 0.7143,
+      "step": 777
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990634183462964,
+      "loss": 0.822,
+      "step": 778
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019906093299711447,
+      "loss": 0.8451,
+      "step": 779
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990584443702477,
+      "loss": 0.7746,
+      "step": 780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019905595246577846,
+      "loss": 0.816,
+      "step": 781
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019905345728378918,
+      "loss": 0.7942,
+      "step": 782
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019905095882436245,
+      "loss": 0.7911,
+      "step": 783
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990484570875809,
+      "loss": 0.8984,
+      "step": 784
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019904595207352737,
+      "loss": 0.8064,
+      "step": 785
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019904344378228467,
+      "loss": 0.8041,
+      "step": 786
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990409322139358,
+      "loss": 0.8366,
+      "step": 787
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019903841736856393,
+      "loss": 0.8618,
+      "step": 788
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019903589924625222,
+      "loss": 0.7576,
+      "step": 789
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019903337784708396,
+      "loss": 0.8154,
+      "step": 790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019903085317114262,
+      "loss": 0.8126,
+      "step": 791
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019902832521851175,
+      "loss": 0.8073,
+      "step": 792
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019902579398927496,
+      "loss": 0.8192,
+      "step": 793
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019902325948351602,
+      "loss": 0.8361,
+      "step": 794
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019902072170131877,
+      "loss": 0.7847,
+      "step": 795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019901818064276724,
+      "loss": 0.8294,
+      "step": 796
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001990156363079454,
+      "loss": 0.7673,
+      "step": 797
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019901308869693754,
+      "loss": 0.7564,
+      "step": 798
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019901053780982787,
+      "loss": 0.8445,
+      "step": 799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001990079836467009,
+      "loss": 0.8028,
+      "step": 800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000199005426207641,
+      "loss": 0.8096,
+      "step": 801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001990028654927329,
+      "loss": 0.7956,
+      "step": 802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019900030150206129,
+      "loss": 0.8012,
+      "step": 803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000198997734235711,
+      "loss": 0.8277,
+      "step": 804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019899516369376697,
+      "loss": 0.7085,
+      "step": 805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989925898763143,
+      "loss": 0.7651,
+      "step": 806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019899001278343806,
+      "loss": 0.7517,
+      "step": 807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989874324152236,
+      "loss": 0.7644,
+      "step": 808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989848487717563,
+      "loss": 0.7515,
+      "step": 809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019898226185312162,
+      "loss": 0.7261,
+      "step": 810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019897967165940513,
+      "loss": 0.7812,
+      "step": 811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989770781906926,
+      "loss": 0.6985,
+      "step": 812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019897448144706975,
+      "loss": 0.8777,
+      "step": 813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989718814286226,
+      "loss": 0.7978,
+      "step": 814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019896927813543708,
+      "loss": 0.7836,
+      "step": 815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019896667156759937,
+      "loss": 0.7953,
+      "step": 816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019896406172519577,
+      "loss": 0.7592,
+      "step": 817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019896144860831257,
+      "loss": 0.7941,
+      "step": 818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019895883221703625,
+      "loss": 0.7828,
+      "step": 819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019895621255145332,
+      "loss": 0.8044,
+      "step": 820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019895358961165055,
+      "loss": 0.8112,
+      "step": 821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019895096339771472,
+      "loss": 0.7647,
+      "step": 822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019894833390973266,
+      "loss": 0.7498,
+      "step": 823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019894570114779142,
+      "loss": 0.761,
+      "step": 824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989430651119781,
+      "loss": 0.8068,
+      "step": 825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019894042580237992,
+      "loss": 0.7604,
+      "step": 826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989377832190842,
+      "loss": 0.7291,
+      "step": 827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989351373621784,
+      "loss": 0.8001,
+      "step": 828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019893248823175003,
+      "loss": 0.7971,
+      "step": 829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019892983582788678,
+      "loss": 0.7617,
+      "step": 830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989271801506764,
+      "loss": 0.7742,
+      "step": 831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019892452120020675,
+      "loss": 0.7633,
+      "step": 832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019892185897656578,
+      "loss": 0.7707,
+      "step": 833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019891919347984164,
+      "loss": 0.7406,
+      "step": 834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989165247101225,
+      "loss": 0.7176,
+      "step": 835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019891385266749667,
+      "loss": 0.7351,
+      "step": 836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019891117735205253,
+      "loss": 0.762,
+      "step": 837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019890849876387861,
+      "loss": 0.7959,
+      "step": 838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019890581690306357,
+      "loss": 0.7431,
+      "step": 839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001989031317696961,
+      "loss": 0.7786,
+      "step": 840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019890044336386508,
+      "loss": 0.8056,
+      "step": 841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019889775168565943,
+      "loss": 0.8553,
+      "step": 842
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019889505673516825,
+      "loss": 0.7481,
+      "step": 843
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019889235851248068,
+      "loss": 0.7106,
+      "step": 844
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000198889657017686,
+      "loss": 0.8818,
+      "step": 845
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019888695225087365,
+      "loss": 0.7524,
+      "step": 846
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019888424421213302,
+      "loss": 0.8297,
+      "step": 847
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001988815329015538,
+      "loss": 0.7312,
+      "step": 848
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019887881831922568,
+      "loss": 0.8348,
+      "step": 849
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019887610046523846,
+      "loss": 0.7978,
+      "step": 850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019887337933968205,
+      "loss": 0.7902,
+      "step": 851
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019887065494264653,
+      "loss": 0.7657,
+      "step": 852
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019886792727422207,
+      "loss": 0.7978,
+      "step": 853
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019886519633449882,
+      "loss": 0.723,
+      "step": 854
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019886246212356723,
+      "loss": 0.7611,
+      "step": 855
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019885972464151773,
+      "loss": 0.7221,
+      "step": 856
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001988569838884409,
+      "loss": 0.7602,
+      "step": 857
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019885423986442744,
+      "loss": 0.7977,
+      "step": 858
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001988514925695681,
+      "loss": 0.7477,
+      "step": 859
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019884874200395384,
+      "loss": 0.7951,
+      "step": 860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019884598816767563,
+      "loss": 0.802,
+      "step": 861
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001988432310608246,
+      "loss": 0.798,
+      "step": 862
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019884047068349198,
+      "loss": 0.7599,
+      "step": 863
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019883770703576912,
+      "loss": 0.7682,
+      "step": 864
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001988349401177474,
+      "loss": 0.8321,
+      "step": 865
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019883216992951845,
+      "loss": 0.7511,
+      "step": 866
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019882939647117387,
+      "loss": 0.7265,
+      "step": 867
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019882661974280542,
+      "loss": 0.7541,
+      "step": 868
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019882383974450507,
+      "loss": 0.7121,
+      "step": 869
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019882105647636472,
+      "loss": 0.7444,
+      "step": 870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019881826993847646,
+      "loss": 0.7698,
+      "step": 871
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001988154801309325,
+      "loss": 0.7572,
+      "step": 872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019881268705382516,
+      "loss": 0.7632,
+      "step": 873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019880989070724688,
+      "loss": 0.7349,
+      "step": 874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019880709109129012,
+      "loss": 0.8271,
+      "step": 875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019880428820604759,
+      "loss": 0.8073,
+      "step": 876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019880148205161192,
+      "loss": 0.7872,
+      "step": 877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001987986726280761,
+      "loss": 0.7895,
+      "step": 878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019879585993553298,
+      "loss": 0.7783,
+      "step": 879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001987930439740757,
+      "loss": 0.767,
+      "step": 880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019879022474379734,
+      "loss": 0.7983,
+      "step": 881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019878740224479126,
+      "loss": 0.832,
+      "step": 882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019878457647715082,
+      "loss": 0.7455,
+      "step": 883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019878174744096953,
+      "loss": 0.8733,
+      "step": 884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019877891513634101,
+      "loss": 0.7894,
+      "step": 885
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019877607956335892,
+      "loss": 0.7341,
+      "step": 886
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019877324072211714,
+      "loss": 0.7487,
+      "step": 887
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019877039861270954,
+      "loss": 0.7991,
+      "step": 888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019876755323523023,
+      "loss": 0.7892,
+      "step": 889
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001987647045897733,
+      "loss": 0.7648,
+      "step": 890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019876185267643303,
+      "loss": 0.8899,
+      "step": 891
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019875899749530377,
+      "loss": 0.7375,
+      "step": 892
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019875613904648,
+      "loss": 0.7643,
+      "step": 893
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001987532773300563,
+      "loss": 0.8038,
+      "step": 894
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019875041234612738,
+      "loss": 0.8279,
+      "step": 895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000198747544094788,
+      "loss": 0.8081,
+      "step": 896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019874467257613305,
+      "loss": 0.7315,
+      "step": 897
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001987417977902576,
+      "loss": 0.7956,
+      "step": 898
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001987389197372567,
+      "loss": 0.7465,
+      "step": 899
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019873603841722565,
+      "loss": 0.7667,
+      "step": 900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001987331538302597,
+      "loss": 0.7133,
+      "step": 901
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019873026597645442,
+      "loss": 0.7452,
+      "step": 902
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001987273748559052,
+      "loss": 0.8016,
+      "step": 903
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019872448046870785,
+      "loss": 0.8022,
+      "step": 904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019872158281495803,
+      "loss": 0.851,
+      "step": 905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001987186818947517,
+      "loss": 0.7805,
+      "step": 906
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019871577770818477,
+      "loss": 0.766,
+      "step": 907
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019871287025535338,
+      "loss": 0.7636,
+      "step": 908
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001987099595363537,
+      "loss": 0.7535,
+      "step": 909
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019870704555128205,
+      "loss": 0.7363,
+      "step": 910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019870412830023487,
+      "loss": 0.7769,
+      "step": 911
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019870120778330867,
+      "loss": 0.7979,
+      "step": 912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019869828400060008,
+      "loss": 0.7649,
+      "step": 913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019869535695220584,
+      "loss": 0.7042,
+      "step": 914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001986924266382228,
+      "loss": 0.8487,
+      "step": 915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001986894930587479,
+      "loss": 0.7587,
+      "step": 916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019868655621387826,
+      "loss": 0.7838,
+      "step": 917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019868361610371097,
+      "loss": 0.7927,
+      "step": 918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001986806727283434,
+      "loss": 0.7852,
+      "step": 919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019867772608787285,
+      "loss": 0.8045,
+      "step": 920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001986747761823969,
+      "loss": 0.8108,
+      "step": 921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019867182301201307,
+      "loss": 0.7717,
+      "step": 922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001986688665768192,
+      "loss": 0.7009,
+      "step": 923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019866590687691294,
+      "loss": 0.8697,
+      "step": 924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019866294391239237,
+      "loss": 0.7736,
+      "step": 925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019865997768335545,
+      "loss": 0.7315,
+      "step": 926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019865700818990034,
+      "loss": 0.8477,
+      "step": 927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019865403543212532,
+      "loss": 0.7362,
+      "step": 928
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001986510594101287,
+      "loss": 0.7611,
+      "step": 929
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019864808012400896,
+      "loss": 0.7781,
+      "step": 930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019864509757386474,
+      "loss": 0.7352,
+      "step": 931
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019864211175979466,
+      "loss": 0.7676,
+      "step": 932
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001986391226818975,
+      "loss": 0.7561,
+      "step": 933
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019863613034027224,
+      "loss": 0.8338,
+      "step": 934
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019863313473501783,
+      "loss": 0.7424,
+      "step": 935
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019863013586623343,
+      "loss": 0.7763,
+      "step": 936
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001986271337340182,
+      "loss": 0.7435,
+      "step": 937
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001986241283384715,
+      "loss": 0.7912,
+      "step": 938
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001986211196796928,
+      "loss": 0.8143,
+      "step": 939
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019861810775778164,
+      "loss": 0.8166,
+      "step": 940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019861509257283766,
+      "loss": 0.8158,
+      "step": 941
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019861207412496062,
+      "loss": 0.8503,
+      "step": 942
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001986090524142504,
+      "loss": 0.7534,
+      "step": 943
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000198606027440807,
+      "loss": 0.7631,
+      "step": 944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019860299920473052,
+      "loss": 0.7763,
+      "step": 945
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001985999677061211,
+      "loss": 0.7873,
+      "step": 946
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019859693294507907,
+      "loss": 0.824,
+      "step": 947
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019859389492170486,
+      "loss": 0.7102,
+      "step": 948
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019859085363609896,
+      "loss": 0.7379,
+      "step": 949
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019858780908836208,
+      "loss": 0.6864,
+      "step": 950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019858476127859485,
+      "loss": 0.7555,
+      "step": 951
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019858171020689818,
+      "loss": 0.8092,
+      "step": 952
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019857865587337295,
+      "loss": 0.7642,
+      "step": 953
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019857559827812034,
+      "loss": 0.732,
+      "step": 954
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019857253742124146,
+      "loss": 0.7683,
+      "step": 955
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019856947330283752,
+      "loss": 0.7734,
+      "step": 956
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019856640592300998,
+      "loss": 0.7151,
+      "step": 957
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019856333528186034,
+      "loss": 0.8122,
+      "step": 958
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019856026137949018,
+      "loss": 0.8087,
+      "step": 959
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019855718421600118,
+      "loss": 0.8221,
+      "step": 960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019855410379149518,
+      "loss": 0.774,
+      "step": 961
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001985510201060741,
+      "loss": 0.8129,
+      "step": 962
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019854793315983996,
+      "loss": 0.7458,
+      "step": 963
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019854484295289492,
+      "loss": 0.8116,
+      "step": 964
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019854174948534124,
+      "loss": 0.7549,
+      "step": 965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019853865275728128,
+      "loss": 0.732,
+      "step": 966
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019853555276881744,
+      "loss": 0.7677,
+      "step": 967
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019853244952005233,
+      "loss": 0.7637,
+      "step": 968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019852934301108863,
+      "loss": 0.7745,
+      "step": 969
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019852623324202916,
+      "loss": 0.7518,
+      "step": 970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019852312021297674,
+      "loss": 0.7796,
+      "step": 971
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019852000392403443,
+      "loss": 0.7407,
+      "step": 972
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001985168843753053,
+      "loss": 0.7489,
+      "step": 973
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019851376156689263,
+      "loss": 0.7414,
+      "step": 974
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001985106354988997,
+      "loss": 0.7586,
+      "step": 975
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001985075061714299,
+      "loss": 0.7892,
+      "step": 976
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019850437358458689,
+      "loss": 0.8236,
+      "step": 977
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019850123773847424,
+      "loss": 0.7523,
+      "step": 978
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019849809863319574,
+      "loss": 0.7911,
+      "step": 979
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019849495626885516,
+      "loss": 0.7481,
+      "step": 980
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019849181064555662,
+      "loss": 0.759,
+      "step": 981
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019848866176340414,
+      "loss": 0.738,
+      "step": 982
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019848550962250184,
+      "loss": 0.7288,
+      "step": 983
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019848235422295412,
+      "loss": 0.7329,
+      "step": 984
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019847919556486532,
+      "loss": 0.7725,
+      "step": 985
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019847603364833998,
+      "loss": 0.7523,
+      "step": 986
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001984728684734827,
+      "loss": 0.7913,
+      "step": 987
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019846970004039824,
+      "loss": 0.8298,
+      "step": 988
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019846652834919142,
+      "loss": 0.7841,
+      "step": 989
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019846335339996715,
+      "loss": 0.7988,
+      "step": 990
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019846017519283056,
+      "loss": 0.7639,
+      "step": 991
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019845699372788672,
+      "loss": 0.8326,
+      "step": 992
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019845380900524094,
+      "loss": 0.7359,
+      "step": 993
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001984506210249986,
+      "loss": 0.7713,
+      "step": 994
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001984474297872652,
+      "loss": 0.8214,
+      "step": 995
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019844423529214627,
+      "loss": 0.7645,
+      "step": 996
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019844103753974753,
+      "loss": 0.7315,
+      "step": 997
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019843783653017484,
+      "loss": 0.7482,
+      "step": 998
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019843463226353408,
+      "loss": 0.8578,
+      "step": 999
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001984314247399312,
+      "loss": 0.7806,
+      "step": 1000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019842821395947246,
+      "loss": 0.6996,
+      "step": 1001
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000198424999922264,
+      "loss": 0.8047,
+      "step": 1002
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001984217826284122,
+      "loss": 0.7923,
+      "step": 1003
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019841856207802349,
+      "loss": 0.7427,
+      "step": 1004
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001984153382712045,
+      "loss": 0.813,
+      "step": 1005
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019841211120806177,
+      "loss": 0.8599,
+      "step": 1006
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019840888088870222,
+      "loss": 0.779,
+      "step": 1007
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001984056473132326,
+      "loss": 0.7752,
+      "step": 1008
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019840241048176002,
+      "loss": 0.7345,
+      "step": 1009
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019839917039439148,
+      "loss": 0.7673,
+      "step": 1010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019839592705123428,
+      "loss": 0.7406,
+      "step": 1011
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019839268045239566,
+      "loss": 0.8254,
+      "step": 1012
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019838943059798304,
+      "loss": 0.7944,
+      "step": 1013
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019838617748810402,
+      "loss": 0.818,
+      "step": 1014
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019838292112286617,
+      "loss": 0.7663,
+      "step": 1015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019837966150237728,
+      "loss": 0.8176,
+      "step": 1016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019837639862674516,
+      "loss": 0.8015,
+      "step": 1017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001983731324960778,
+      "loss": 0.795,
+      "step": 1018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019836986311048326,
+      "loss": 0.8265,
+      "step": 1019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019836659047006973,
+      "loss": 0.7223,
+      "step": 1020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019836331457494545,
+      "loss": 0.7249,
+      "step": 1021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019836003542521887,
+      "loss": 0.7122,
+      "step": 1022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019835675302099848,
+      "loss": 0.7813,
+      "step": 1023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019835346736239284,
+      "loss": 0.7657,
+      "step": 1024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001983501784495107,
+      "loss": 0.7757,
+      "step": 1025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019834688628246085,
+      "loss": 0.774,
+      "step": 1026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019834359086135227,
+      "loss": 0.7902,
+      "step": 1027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000198340292186294,
+      "loss": 0.7149,
+      "step": 1028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001983369902573951,
+      "loss": 0.7779,
+      "step": 1029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001983336850747649,
+      "loss": 0.7739,
+      "step": 1030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019833037663851276,
+      "loss": 0.7903,
+      "step": 1031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001983270649487481,
+      "loss": 0.7982,
+      "step": 1032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019832375000558054,
+      "loss": 0.7192,
+      "step": 1033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019832043180911975,
+      "loss": 0.7684,
+      "step": 1034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001983171103594755,
+      "loss": 0.7968,
+      "step": 1035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019831378565675773,
+      "loss": 0.8108,
+      "step": 1036
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001983104577010764,
+      "loss": 0.8191,
+      "step": 1037
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019830712649254167,
+      "loss": 0.7044,
+      "step": 1038
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019830379203126375,
+      "loss": 0.7365,
+      "step": 1039
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019830045431735294,
+      "loss": 0.8285,
+      "step": 1040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001982971133509197,
+      "loss": 0.7784,
+      "step": 1041
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019829376913207456,
+      "loss": 0.7808,
+      "step": 1042
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019829042166092818,
+      "loss": 0.7859,
+      "step": 1043
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019828707093759134,
+      "loss": 0.7109,
+      "step": 1044
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019828371696217488,
+      "loss": 0.8352,
+      "step": 1045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001982803597347898,
+      "loss": 0.7537,
+      "step": 1046
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001982769992555471,
+      "loss": 0.7226,
+      "step": 1047
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001982736355245581,
+      "loss": 0.7469,
+      "step": 1048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019827026854193403,
+      "loss": 0.7409,
+      "step": 1049
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019826689830778626,
+      "loss": 0.7756,
+      "step": 1050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019826352482222638,
+      "loss": 0.7978,
+      "step": 1051
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019826014808536595,
+      "loss": 0.8058,
+      "step": 1052
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019825676809731671,
+      "loss": 0.7949,
+      "step": 1053
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019825338485819055,
+      "loss": 0.8137,
+      "step": 1054
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019824999836809935,
+      "loss": 0.7905,
+      "step": 1055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019824660862715517,
+      "loss": 0.8283,
+      "step": 1056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001982432156354702,
+      "loss": 0.7833,
+      "step": 1057
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00019823981939315664,
+      "loss": 0.744,
+      "step": 1058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019823641990032693,
+      "loss": 0.7744,
+      "step": 1059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019823301715709356,
+      "loss": 0.8207,
+      "step": 1060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019822961116356904,
+      "loss": 0.7677,
+      "step": 1061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001982262019198661,
+      "loss": 0.7826,
+      "step": 1062
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001982227894260976,
+      "loss": 0.8715,
+      "step": 1063
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001982193736823764,
+      "loss": 0.8244,
+      "step": 1064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001982159546888155,
+      "loss": 0.8314,
+      "step": 1065
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001982125324455281,
+      "loss": 0.7528,
+      "step": 1066
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019820910695262734,
+      "loss": 0.7973,
+      "step": 1067
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019820567821022663,
+      "loss": 0.773,
+      "step": 1068
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019820224621843938,
+      "loss": 0.7946,
+      "step": 1069
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019819881097737915,
+      "loss": 0.7557,
+      "step": 1070
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019819537248715966,
+      "loss": 0.7288,
+      "step": 1071
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981919307478946,
+      "loss": 0.7538,
+      "step": 1072
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981884857596979,
+      "loss": 0.8472,
+      "step": 1073
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981850375226835,
+      "loss": 0.7633,
+      "step": 1074
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019818158603696558,
+      "loss": 0.7423,
+      "step": 1075
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019817813130265825,
+      "loss": 0.7788,
+      "step": 1076
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019817467331987586,
+      "loss": 0.6996,
+      "step": 1077
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019817121208873283,
+      "loss": 0.7042,
+      "step": 1078
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019816774760934367,
+      "loss": 0.8109,
+      "step": 1079
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019816427988182304,
+      "loss": 0.7844,
+      "step": 1080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019816080890628564,
+      "loss": 0.8378,
+      "step": 1081
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019815733468284632,
+      "loss": 0.8074,
+      "step": 1082
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019815385721162003,
+      "loss": 0.8401,
+      "step": 1083
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981503764927219,
+      "loss": 0.7465,
+      "step": 1084
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000198146892526267,
+      "loss": 0.7957,
+      "step": 1085
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019814340531237065,
+      "loss": 0.7814,
+      "step": 1086
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019813991485114824,
+      "loss": 0.7633,
+      "step": 1087
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019813642114271528,
+      "loss": 0.7503,
+      "step": 1088
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019813292418718732,
+      "loss": 0.8464,
+      "step": 1089
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981294239846801,
+      "loss": 0.738,
+      "step": 1090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019812592053530944,
+      "loss": 0.7812,
+      "step": 1091
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981224138391912,
+      "loss": 0.8099,
+      "step": 1092
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019811890389644145,
+      "loss": 0.7437,
+      "step": 1093
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019811539070717634,
+      "loss": 0.787,
+      "step": 1094
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019811187427151212,
+      "loss": 0.8305,
+      "step": 1095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001981083545895651,
+      "loss": 0.7848,
+      "step": 1096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019810483166145178,
+      "loss": 0.7131,
+      "step": 1097
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019810130548728866,
+      "loss": 0.7572,
+      "step": 1098
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001980977760671925,
+      "loss": 0.7203,
+      "step": 1099
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019809424340127998,
+      "loss": 0.7462,
+      "step": 1100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019809070748966808,
+      "loss": 0.7526,
+      "step": 1101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019808716833247375,
+      "loss": 0.8209,
+      "step": 1102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019808362592981408,
+      "loss": 0.765,
+      "step": 1103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019808008028180632,
+      "loss": 0.7736,
+      "step": 1104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001980765313885678,
+      "loss": 0.798,
+      "step": 1105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019807297925021584,
+      "loss": 0.7512,
+      "step": 1106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019806942386686805,
+      "loss": 0.7517,
+      "step": 1107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001980658652386421,
+      "loss": 0.7331,
+      "step": 1108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019806230336565566,
+      "loss": 0.6801,
+      "step": 1109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019805873824802666,
+      "loss": 0.7871,
+      "step": 1110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000198055169885873,
+      "loss": 0.8183,
+      "step": 1111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019805159827931276,
+      "loss": 0.7986,
+      "step": 1112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019804802342846412,
+      "loss": 0.8222,
+      "step": 1113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001980444453334454,
+      "loss": 0.7497,
+      "step": 1114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019804086399437493,
+      "loss": 0.7949,
+      "step": 1115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019803727941137124,
+      "loss": 0.8546,
+      "step": 1116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019803369158455294,
+      "loss": 0.7361,
+      "step": 1117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019803010051403872,
+      "loss": 0.8341,
+      "step": 1118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019802650619994743,
+      "loss": 0.7617,
+      "step": 1119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019802290864239796,
+      "loss": 0.7573,
+      "step": 1120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001980193078415094,
+      "loss": 0.799,
+      "step": 1121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019801570379740082,
+      "loss": 0.7718,
+      "step": 1122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019801209651019154,
+      "loss": 0.7355,
+      "step": 1123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019800848598000082,
+      "loss": 0.818,
+      "step": 1124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019800487220694826,
+      "loss": 0.7413,
+      "step": 1125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019800125519115331,
+      "loss": 0.8442,
+      "step": 1126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001979976349327357,
+      "loss": 0.7562,
+      "step": 1127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019799401143181524,
+      "loss": 0.7964,
+      "step": 1128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019799038468851178,
+      "loss": 0.7697,
+      "step": 1129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019798675470294533,
+      "loss": 0.815,
+      "step": 1130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019798312147523598,
+      "loss": 0.7861,
+      "step": 1131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019797948500550398,
+      "loss": 0.7241,
+      "step": 1132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019797584529386962,
+      "loss": 0.7592,
+      "step": 1133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001979722023404534,
+      "loss": 0.7846,
+      "step": 1134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019796855614537576,
+      "loss": 0.8187,
+      "step": 1135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001979649067087574,
+      "loss": 0.645,
+      "step": 1136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019796125403071907,
+      "loss": 0.7446,
+      "step": 1137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019795759811138156,
+      "loss": 0.7413,
+      "step": 1138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019795393895086596,
+      "loss": 0.6966,
+      "step": 1139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019795027654929325,
+      "loss": 0.7282,
+      "step": 1140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019794661090678464,
+      "loss": 0.8217,
+      "step": 1141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001979429420234614,
+      "loss": 0.7984,
+      "step": 1142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019793926989944495,
+      "loss": 0.8325,
+      "step": 1143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019793559453485675,
+      "loss": 0.8431,
+      "step": 1144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019793191592981844,
+      "loss": 0.7888,
+      "step": 1145
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019792823408445174,
+      "loss": 0.7064,
+      "step": 1146
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019792454899887843,
+      "loss": 0.7815,
+      "step": 1147
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019792086067322052,
+      "loss": 0.7514,
+      "step": 1148
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001979171691076,
+      "loss": 0.7511,
+      "step": 1149
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019791347430213904,
+      "loss": 0.7879,
+      "step": 1150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019790977625695985,
+      "loss": 0.713,
+      "step": 1151
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001979060749721848,
+      "loss": 0.7451,
+      "step": 1152
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019790237044793637,
+      "loss": 0.8109,
+      "step": 1153
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019789866268433716,
+      "loss": 0.7597,
+      "step": 1154
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019789495168150975,
+      "loss": 0.7814,
+      "step": 1155
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019789123743957707,
+      "loss": 0.762,
+      "step": 1156
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001978875199586619,
+      "loss": 0.7868,
+      "step": 1157
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001978837992388873,
+      "loss": 0.7238,
+      "step": 1158
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019788007528037638,
+      "loss": 0.8193,
+      "step": 1159
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019787634808325234,
+      "loss": 0.8377,
+      "step": 1160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001978726176476385,
+      "loss": 0.8235,
+      "step": 1161
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001978688839736583,
+      "loss": 0.7345,
+      "step": 1162
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019786514706143527,
+      "loss": 0.7323,
+      "step": 1163
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019786140691109306,
+      "loss": 0.7857,
+      "step": 1164
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019785766352275542,
+      "loss": 0.7616,
+      "step": 1165
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019785391689654621,
+      "loss": 0.7761,
+      "step": 1166
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001978501670325894,
+      "loss": 0.7486,
+      "step": 1167
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001978464139310091,
+      "loss": 0.7986,
+      "step": 1168
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019784265759192944,
+      "loss": 0.7937,
+      "step": 1169
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001978388980154747,
+      "loss": 0.8125,
+      "step": 1170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019783513520176931,
+      "loss": 0.6969,
+      "step": 1171
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019783136915093775,
+      "loss": 0.7992,
+      "step": 1172
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019782759986310464,
+      "loss": 0.7196,
+      "step": 1173
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019782382733839473,
+      "loss": 0.7758,
+      "step": 1174
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001978200515769328,
+      "loss": 0.8438,
+      "step": 1175
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019781627257884376,
+      "loss": 0.72,
+      "step": 1176
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019781249034425268,
+      "loss": 0.816,
+      "step": 1177
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019780870487328474,
+      "loss": 0.8128,
+      "step": 1178
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019780491616606512,
+      "loss": 0.814,
+      "step": 1179
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019780112422271927,
+      "loss": 0.7655,
+      "step": 1180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019779732904337254,
+      "loss": 0.6974,
+      "step": 1181
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019779353062815059,
+      "loss": 0.7732,
+      "step": 1182
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019778972897717908,
+      "loss": 0.739,
+      "step": 1183
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019778592409058378,
+      "loss": 0.749,
+      "step": 1184
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019778211596849059,
+      "loss": 0.7604,
+      "step": 1185
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019777830461102552,
+      "loss": 0.7535,
+      "step": 1186
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019777449001831467,
+      "loss": 0.7586,
+      "step": 1187
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977706721904843,
+      "loss": 0.757,
+      "step": 1188
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019776685112766064,
+      "loss": 0.7932,
+      "step": 1189
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977630268299702,
+      "loss": 0.7751,
+      "step": 1190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019775919929753953,
+      "loss": 0.7104,
+      "step": 1191
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977553685304952,
+      "loss": 0.7446,
+      "step": 1192
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019775153452896398,
+      "loss": 0.7957,
+      "step": 1193
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019774769729307274,
+      "loss": 0.7605,
+      "step": 1194
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019774385682294848,
+      "loss": 0.7869,
+      "step": 1195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019774001311871823,
+      "loss": 0.7699,
+      "step": 1196
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019773616618050916,
+      "loss": 0.7761,
+      "step": 1197
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977323160084486,
+      "loss": 0.804,
+      "step": 1198
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977284626026639,
+      "loss": 0.7682,
+      "step": 1199
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977246059632826,
+      "loss": 0.7562,
+      "step": 1200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019772074609043222,
+      "loss": 0.7447,
+      "step": 1201
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977168829842406,
+      "loss": 0.7559,
+      "step": 1202
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977130166448355,
+      "loss": 0.7783,
+      "step": 1203
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977091470723448,
+      "loss": 0.7547,
+      "step": 1204
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001977052742668966,
+      "loss": 0.7868,
+      "step": 1205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019770139822861905,
+      "loss": 0.7915,
+      "step": 1206
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019769751895764036,
+      "loss": 0.7113,
+      "step": 1207
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019769363645408887,
+      "loss": 0.6996,
+      "step": 1208
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019768975071809307,
+      "loss": 0.8125,
+      "step": 1209
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019768586174978157,
+      "loss": 0.769,
+      "step": 1210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019768196954928296,
+      "loss": 0.8157,
+      "step": 1211
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001976780741167261,
+      "loss": 0.8269,
+      "step": 1212
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019767417545223982,
+      "loss": 0.7532,
+      "step": 1213
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019767027355595317,
+      "loss": 0.81,
+      "step": 1214
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001976663684279952,
+      "loss": 0.7285,
+      "step": 1215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019766246006849522,
+      "loss": 0.7789,
+      "step": 1216
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001976585484775824,
+      "loss": 0.768,
+      "step": 1217
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001976546336553863,
+      "loss": 0.7111,
+      "step": 1218
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019765071560203637,
+      "loss": 0.7495,
+      "step": 1219
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001976467943176623,
+      "loss": 0.8338,
+      "step": 1220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019764286980239378,
+      "loss": 0.7826,
+      "step": 1221
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019763894205636072,
+      "loss": 0.7478,
+      "step": 1222
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019763501107969303,
+      "loss": 0.7788,
+      "step": 1223
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019763107687252083,
+      "loss": 0.8376,
+      "step": 1224
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019762713943497425,
+      "loss": 0.7509,
+      "step": 1225
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019762319876718355,
+      "loss": 0.7513,
+      "step": 1226
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001976192548692792,
+      "loss": 0.7572,
+      "step": 1227
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001976153077413916,
+      "loss": 0.7489,
+      "step": 1228
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019761135738365142,
+      "loss": 0.6974,
+      "step": 1229
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019760740379618934,
+      "loss": 0.7612,
+      "step": 1230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019760344697913618,
+      "loss": 0.7539,
+      "step": 1231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019759948693262287,
+      "loss": 0.7993,
+      "step": 1232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001975955236567804,
+      "loss": 0.7668,
+      "step": 1233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019759155715173997,
+      "loss": 0.7406,
+      "step": 1234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019758758741763276,
+      "loss": 0.84,
+      "step": 1235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019758361445459014,
+      "loss": 0.8156,
+      "step": 1236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019757963826274357,
+      "loss": 0.8152,
+      "step": 1237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001975756588422246,
+      "loss": 0.7422,
+      "step": 1238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019757167619316495,
+      "loss": 0.7926,
+      "step": 1239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019756769031569632,
+      "loss": 0.7889,
+      "step": 1240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019756370120995066,
+      "loss": 0.7503,
+      "step": 1241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001975597088760599,
+      "loss": 0.7746,
+      "step": 1242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001975557133141562,
+      "loss": 0.7578,
+      "step": 1243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001975517145243717,
+      "loss": 0.7968,
+      "step": 1244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019754771250683875,
+      "loss": 0.835,
+      "step": 1245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019754370726168973,
+      "loss": 0.8047,
+      "step": 1246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019753969878905724,
+      "loss": 0.7857,
+      "step": 1247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019753568708907383,
+      "loss": 0.8333,
+      "step": 1248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001975316721618723,
+      "loss": 0.7447,
+      "step": 1249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019752765400758544,
+      "loss": 0.71,
+      "step": 1250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001975236326263462,
+      "loss": 0.744,
+      "step": 1251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019751960801828769,
+      "loss": 0.7605,
+      "step": 1252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019751558018354305,
+      "loss": 0.8101,
+      "step": 1253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019751154912224555,
+      "loss": 0.7548,
+      "step": 1254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019750751483452853,
+      "loss": 0.7587,
+      "step": 1255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019750347732052555,
+      "loss": 0.7774,
+      "step": 1256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019749943658037012,
+      "loss": 0.7791,
+      "step": 1257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019749539261419602,
+      "loss": 0.7933,
+      "step": 1258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000197491345422137,
+      "loss": 0.738,
+      "step": 1259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000197487295004327,
+      "loss": 0.7631,
+      "step": 1260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001974832413609,
+      "loss": 0.7455,
+      "step": 1261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019747918449199015,
+      "loss": 0.774,
+      "step": 1262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019747512439773171,
+      "loss": 0.7756,
+      "step": 1263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019747106107825897,
+      "loss": 0.8287,
+      "step": 1264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019746699453370638,
+      "loss": 0.7537,
+      "step": 1265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019746292476420854,
+      "loss": 0.7651,
+      "step": 1266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019745885176990007,
+      "loss": 0.7309,
+      "step": 1267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019745477555091574,
+      "loss": 0.7243,
+      "step": 1268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001974506961073904,
+      "loss": 0.7385,
+      "step": 1269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019744661343945908,
+      "loss": 0.7365,
+      "step": 1270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019744252754725684,
+      "loss": 0.7134,
+      "step": 1271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019743843843091888,
+      "loss": 0.8651,
+      "step": 1272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019743434609058045,
+      "loss": 0.7229,
+      "step": 1273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019743025052637703,
+      "loss": 0.7968,
+      "step": 1274
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019742615173844405,
+      "loss": 0.7774,
+      "step": 1275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019742204972691724,
+      "loss": 0.7085,
+      "step": 1276
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019741794449193223,
+      "loss": 0.7432,
+      "step": 1277
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019741383603362489,
+      "loss": 0.7992,
+      "step": 1278
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019740972435213115,
+      "loss": 0.7319,
+      "step": 1279
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019740560944758704,
+      "loss": 0.7213,
+      "step": 1280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019740149132012875,
+      "loss": 0.682,
+      "step": 1281
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019739736996989254,
+      "loss": 0.7608,
+      "step": 1282
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019739324539701475,
+      "loss": 0.7845,
+      "step": 1283
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019738911760163182,
+      "loss": 0.7424,
+      "step": 1284
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001973849865838804,
+      "loss": 0.7311,
+      "step": 1285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001973808523438971,
+      "loss": 0.8003,
+      "step": 1286
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001973767148818188,
+      "loss": 0.738,
+      "step": 1287
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019737257419778233,
+      "loss": 0.8216,
+      "step": 1288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019736843029192472,
+      "loss": 0.759,
+      "step": 1289
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019736428316438306,
+      "loss": 0.7788,
+      "step": 1290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019736013281529462,
+      "loss": 0.722,
+      "step": 1291
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019735597924479668,
+      "loss": 0.7488,
+      "step": 1292
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019735182245302668,
+      "loss": 0.7334,
+      "step": 1293
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019734766244012216,
+      "loss": 0.7657,
+      "step": 1294
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019734349920622076,
+      "loss": 0.766,
+      "step": 1295
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019733933275146027,
+      "loss": 0.8192,
+      "step": 1296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019733516307597848,
+      "loss": 0.8064,
+      "step": 1297
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019733099017991341,
+      "loss": 0.7609,
+      "step": 1298
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001973268140634031,
+      "loss": 0.7311,
+      "step": 1299
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019732263472658574,
+      "loss": 0.7082,
+      "step": 1300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001973184521695996,
+      "loss": 0.7526,
+      "step": 1301
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001973142663925831,
+      "loss": 0.7706,
+      "step": 1302
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001973100773956747,
+      "loss": 0.7497,
+      "step": 1303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019730588517901304,
+      "loss": 0.7764,
+      "step": 1304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001973016897427368,
+      "loss": 0.805,
+      "step": 1305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019729749108698483,
+      "loss": 0.8075,
+      "step": 1306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019729328921189602,
+      "loss": 0.7968,
+      "step": 1307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019728908411760943,
+      "loss": 0.7523,
+      "step": 1308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019728487580426417,
+      "loss": 0.7685,
+      "step": 1309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001972806642719995,
+      "loss": 0.7312,
+      "step": 1310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019727644952095475,
+      "loss": 0.7998,
+      "step": 1311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019727223155126936,
+      "loss": 0.8208,
+      "step": 1312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000197268010363083,
+      "loss": 0.7416,
+      "step": 1313
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019726378595653516,
+      "loss": 0.7824,
+      "step": 1314
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019725955833176578,
+      "loss": 0.7247,
+      "step": 1315
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019725532748891467,
+      "loss": 0.753,
+      "step": 1316
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001972510934281218,
+      "loss": 0.7493,
+      "step": 1317
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019724685614952732,
+      "loss": 0.7539,
+      "step": 1318
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019724261565327137,
+      "loss": 0.7477,
+      "step": 1319
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001972383719394943,
+      "loss": 0.7503,
+      "step": 1320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019723412500833654,
+      "loss": 0.6795,
+      "step": 1321
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019722987485993855,
+      "loss": 0.7709,
+      "step": 1322
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000197225621494441,
+      "loss": 0.8006,
+      "step": 1323
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001972213649119846,
+      "loss": 0.8048,
+      "step": 1324
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019721710511271022,
+      "loss": 0.7671,
+      "step": 1325
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019721284209675877,
+      "loss": 0.7418,
+      "step": 1326
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019720857586427135,
+      "loss": 0.8054,
+      "step": 1327
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001972043064153891,
+      "loss": 0.776,
+      "step": 1328
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001972000337502532,
+      "loss": 0.8358,
+      "step": 1329
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019719575786900518,
+      "loss": 0.7456,
+      "step": 1330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019719147877178642,
+      "loss": 0.7653,
+      "step": 1331
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019718719645873853,
+      "loss": 0.7642,
+      "step": 1332
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019718291093000318,
+      "loss": 0.7729,
+      "step": 1333
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001971786221857222,
+      "loss": 0.7621,
+      "step": 1334
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019717433022603745,
+      "loss": 0.7367,
+      "step": 1335
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019717003505109095,
+      "loss": 0.7545,
+      "step": 1336
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001971657366610249,
+      "loss": 0.7744,
+      "step": 1337
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001971614350559814,
+      "loss": 0.8163,
+      "step": 1338
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019715713023610282,
+      "loss": 0.7044,
+      "step": 1339
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019715282220153166,
+      "loss": 0.7756,
+      "step": 1340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001971485109524104,
+      "loss": 0.7794,
+      "step": 1341
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001971441964888817,
+      "loss": 0.7407,
+      "step": 1342
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019713987881108834,
+      "loss": 0.8074,
+      "step": 1343
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019713555791917313,
+      "loss": 0.6788,
+      "step": 1344
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001971312338132791,
+      "loss": 0.8079,
+      "step": 1345
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019712690649354925,
+      "loss": 0.6929,
+      "step": 1346
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019712257596012683,
+      "loss": 0.7825,
+      "step": 1347
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019711824221315508,
+      "loss": 0.8501,
+      "step": 1348
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019711390525277742,
+      "loss": 0.7092,
+      "step": 1349
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019710956507913737,
+      "loss": 0.7794,
+      "step": 1350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019710522169237845,
+      "loss": 0.8125,
+      "step": 1351
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019710087509264447,
+      "loss": 0.7366,
+      "step": 1352
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001970965252800792,
+      "loss": 0.7414,
+      "step": 1353
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001970921722548266,
+      "loss": 0.6955,
+      "step": 1354
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019708781601703065,
+      "loss": 0.8299,
+      "step": 1355
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001970834565668355,
+      "loss": 0.7265,
+      "step": 1356
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001970790939043854,
+      "loss": 0.7327,
+      "step": 1357
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019707472802982473,
+      "loss": 0.7936,
+      "step": 1358
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001970703589432979,
+      "loss": 0.6816,
+      "step": 1359
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00019706598664494948,
+      "loss": 0.7409,
+      "step": 1360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019706161113492416,
+      "loss": 0.799,
+      "step": 1361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019705723241336673,
+      "loss": 0.7756,
+      "step": 1362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019705285048042203,
+      "loss": 0.74,
+      "step": 1363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019704846533623508,
+      "loss": 0.8041,
+      "step": 1364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019704407698095094,
+      "loss": 0.7341,
+      "step": 1365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019703968541471485,
+      "loss": 0.7252,
+      "step": 1366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019703529063767208,
+      "loss": 0.7471,
+      "step": 1367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019703089264996805,
+      "loss": 0.7147,
+      "step": 1368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001970264914517483,
+      "loss": 0.7906,
+      "step": 1369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019702208704315845,
+      "loss": 0.7212,
+      "step": 1370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019701767942434422,
+      "loss": 0.7825,
+      "step": 1371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019701326859545141,
+      "loss": 0.7739,
+      "step": 1372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019700885455662605,
+      "loss": 0.8377,
+      "step": 1373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019700443730801413,
+      "loss": 0.77,
+      "step": 1374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019700001684976178,
+      "loss": 0.7839,
+      "step": 1375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019699559318201538,
+      "loss": 0.7226,
+      "step": 1376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019699116630492115,
+      "loss": 0.7835,
+      "step": 1377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019698673621862566,
+      "loss": 0.7649,
+      "step": 1378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001969823029232755,
+      "loss": 0.7798,
+      "step": 1379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019697786641901724,
+      "loss": 0.8101,
+      "step": 1380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019697342670599782,
+      "loss": 0.7697,
+      "step": 1381
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019696898378436406,
+      "loss": 0.7613,
+      "step": 1382
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019696453765426298,
+      "loss": 0.7521,
+      "step": 1383
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019696008831584168,
+      "loss": 0.717,
+      "step": 1384
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001969556357692474,
+      "loss": 0.7397,
+      "step": 1385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019695118001462745,
+      "loss": 0.761,
+      "step": 1386
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019694672105212925,
+      "loss": 0.8108,
+      "step": 1387
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019694225888190037,
+      "loss": 0.7332,
+      "step": 1388
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019693779350408842,
+      "loss": 0.7841,
+      "step": 1389
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019693332491884117,
+      "loss": 0.753,
+      "step": 1390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019692885312630644,
+      "loss": 0.7629,
+      "step": 1391
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019692437812663226,
+      "loss": 0.7558,
+      "step": 1392
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019691989991996663,
+      "loss": 0.7647,
+      "step": 1393
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019691541850645771,
+      "loss": 0.7704,
+      "step": 1394
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019691093388625385,
+      "loss": 0.7681,
+      "step": 1395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001969064460595034,
+      "loss": 0.7921,
+      "step": 1396
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019690195502635485,
+      "loss": 0.7044,
+      "step": 1397
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001968974607869568,
+      "loss": 0.766,
+      "step": 1398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019689296334145795,
+      "loss": 0.7272,
+      "step": 1399
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019688846269000712,
+      "loss": 0.7544,
+      "step": 1400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019688395883275318,
+      "loss": 0.7985,
+      "step": 1401
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001968794517698452,
+      "loss": 0.8063,
+      "step": 1402
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019687494150143233,
+      "loss": 0.7324,
+      "step": 1403
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019687042802766375,
+      "loss": 0.7604,
+      "step": 1404
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019686591134868882,
+      "loss": 0.7516,
+      "step": 1405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019686139146465697,
+      "loss": 0.8407,
+      "step": 1406
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019685686837571777,
+      "loss": 0.7913,
+      "step": 1407
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019685234208202088,
+      "loss": 0.7138,
+      "step": 1408
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019684781258371607,
+      "loss": 0.7809,
+      "step": 1409
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001968432798809532,
+      "loss": 0.7334,
+      "step": 1410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019683874397388222,
+      "loss": 0.713,
+      "step": 1411
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019683420486265327,
+      "loss": 0.7386,
+      "step": 1412
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019682966254741652,
+      "loss": 0.734,
+      "step": 1413
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019682511702832217,
+      "loss": 0.7865,
+      "step": 1414
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019682056830552078,
+      "loss": 0.7651,
+      "step": 1415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019681601637916275,
+      "loss": 0.7903,
+      "step": 1416
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001968114612493987,
+      "loss": 0.7277,
+      "step": 1417
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019680690291637939,
+      "loss": 0.8142,
+      "step": 1418
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019680234138025562,
+      "loss": 0.7798,
+      "step": 1419
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001967977766411783,
+      "loss": 0.7833,
+      "step": 1420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019679320869929852,
+      "loss": 0.8103,
+      "step": 1421
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019678863755476735,
+      "loss": 0.7925,
+      "step": 1422
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019678406320773612,
+      "loss": 0.7628,
+      "step": 1423
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019677948565835612,
+      "loss": 0.7531,
+      "step": 1424
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001967749049067788,
+      "loss": 0.7372,
+      "step": 1425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001967703209531558,
+      "loss": 0.7487,
+      "step": 1426
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019676573379763873,
+      "loss": 0.789,
+      "step": 1427
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001967611434403794,
+      "loss": 0.7966,
+      "step": 1428
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019675654988152969,
+      "loss": 0.7867,
+      "step": 1429
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019675195312124152,
+      "loss": 0.7119,
+      "step": 1430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001967473531596671,
+      "loss": 0.7759,
+      "step": 1431
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019674274999695858,
+      "loss": 0.7277,
+      "step": 1432
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019673814363326822,
+      "loss": 0.7888,
+      "step": 1433
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019673353406874852,
+      "loss": 0.8067,
+      "step": 1434
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019672892130355194,
+      "loss": 0.7915,
+      "step": 1435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019672430533783112,
+      "loss": 0.7936,
+      "step": 1436
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001967196861717388,
+      "loss": 0.7768,
+      "step": 1437
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019671506380542782,
+      "loss": 0.8452,
+      "step": 1438
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001967104382390511,
+      "loss": 0.6845,
+      "step": 1439
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019670580947276167,
+      "loss": 0.7556,
+      "step": 1440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019670117750671278,
+      "loss": 0.7853,
+      "step": 1441
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019669654234105756,
+      "loss": 0.7619,
+      "step": 1442
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001966919039759495,
+      "loss": 0.8179,
+      "step": 1443
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019668726241154198,
+      "loss": 0.7443,
+      "step": 1444
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019668261764798863,
+      "loss": 0.7861,
+      "step": 1445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001966779696854431,
+      "loss": 0.7287,
+      "step": 1446
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019667331852405923,
+      "loss": 0.7648,
+      "step": 1447
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019666866416399086,
+      "loss": 0.7757,
+      "step": 1448
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019666400660539203,
+      "loss": 0.8381,
+      "step": 1449
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019665934584841682,
+      "loss": 0.7835,
+      "step": 1450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019665468189321943,
+      "loss": 0.7514,
+      "step": 1451
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019665001473995425,
+      "loss": 0.7568,
+      "step": 1452
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019664534438877562,
+      "loss": 0.7476,
+      "step": 1453
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001966406708398381,
+      "loss": 0.7273,
+      "step": 1454
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019663599409329639,
+      "loss": 0.7446,
+      "step": 1455
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019663131414930513,
+      "loss": 0.7291,
+      "step": 1456
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019662663100801923,
+      "loss": 0.6979,
+      "step": 1457
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019662194466959362,
+      "loss": 0.7757,
+      "step": 1458
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001966172551341834,
+      "loss": 0.759,
+      "step": 1459
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019661256240194365,
+      "loss": 0.7731,
+      "step": 1460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019660786647302973,
+      "loss": 0.8103,
+      "step": 1461
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000196603167347597,
+      "loss": 0.8015,
+      "step": 1462
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019659846502580088,
+      "loss": 0.7938,
+      "step": 1463
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019659375950779705,
+      "loss": 0.7421,
+      "step": 1464
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001965890507937411,
+      "loss": 0.7075,
+      "step": 1465
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001965843388837889,
+      "loss": 0.7841,
+      "step": 1466
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019657962377809634,
+      "loss": 0.7959,
+      "step": 1467
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019657490547681948,
+      "loss": 0.6756,
+      "step": 1468
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019657018398011434,
+      "loss": 0.7712,
+      "step": 1469
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019656545928813723,
+      "loss": 0.7499,
+      "step": 1470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019656073140104444,
+      "loss": 0.7739,
+      "step": 1471
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019655600031899238,
+      "loss": 0.7634,
+      "step": 1472
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019655126604213762,
+      "loss": 0.7995,
+      "step": 1473
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001965465285706368,
+      "loss": 0.796,
+      "step": 1474
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001965417879046467,
+      "loss": 0.7791,
+      "step": 1475
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019653704404432414,
+      "loss": 0.804,
+      "step": 1476
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001965322969898261,
+      "loss": 0.7635,
+      "step": 1477
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019652754674130959,
+      "loss": 0.8199,
+      "step": 1478
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001965227932989319,
+      "loss": 0.7179,
+      "step": 1479
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001965180366628502,
+      "loss": 0.8095,
+      "step": 1480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019651327683322195,
+      "loss": 0.743,
+      "step": 1481
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019650851381020458,
+      "loss": 0.6996,
+      "step": 1482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019650374759395576,
+      "loss": 0.8009,
+      "step": 1483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019649897818463312,
+      "loss": 0.8278,
+      "step": 1484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001964942055823945,
+      "loss": 0.7486,
+      "step": 1485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019648942978739783,
+      "loss": 0.8181,
+      "step": 1486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001964846507998011,
+      "loss": 0.7577,
+      "step": 1487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019647986861976246,
+      "loss": 0.7711,
+      "step": 1488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019647508324744013,
+      "loss": 0.8412,
+      "step": 1489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019647029468299244,
+      "loss": 0.7524,
+      "step": 1490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019646550292657785,
+      "loss": 0.7719,
+      "step": 1491
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019646070797835484,
+      "loss": 0.7645,
+      "step": 1492
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019645590983848217,
+      "loss": 0.7755,
+      "step": 1493
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001964511085071185,
+      "loss": 0.8164,
+      "step": 1494
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019644630398442278,
+      "loss": 0.7778,
+      "step": 1495
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019644149627055395,
+      "loss": 0.7815,
+      "step": 1496
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019643668536567106,
+      "loss": 0.7865,
+      "step": 1497
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001964318712699333,
+      "loss": 0.7047,
+      "step": 1498
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019642705398349998,
+      "loss": 0.7686,
+      "step": 1499
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019642223350653043,
+      "loss": 0.7665,
+      "step": 1500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019641740983918426,
+      "loss": 0.7934,
+      "step": 1501
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019641258298162096,
+      "loss": 0.7591,
+      "step": 1502
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001964077529340003,
+      "loss": 0.8176,
+      "step": 1503
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001964029196964821,
+      "loss": 0.7002,
+      "step": 1504
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019639808326922623,
+      "loss": 0.7692,
+      "step": 1505
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019639324365239275,
+      "loss": 0.7874,
+      "step": 1506
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019638840084614182,
+      "loss": 0.7607,
+      "step": 1507
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963835548506336,
+      "loss": 0.7023,
+      "step": 1508
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963787056660285,
+      "loss": 0.7638,
+      "step": 1509
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019637385329248695,
+      "loss": 0.8114,
+      "step": 1510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019636899773016947,
+      "loss": 0.7821,
+      "step": 1511
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963641389792368,
+      "loss": 0.8403,
+      "step": 1512
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963592770398496,
+      "loss": 0.776,
+      "step": 1513
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963544119121688,
+      "loss": 0.7163,
+      "step": 1514
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963495435963554,
+      "loss": 0.7579,
+      "step": 1515
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963446720925704,
+      "loss": 0.7354,
+      "step": 1516
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019633979740097502,
+      "loss": 0.7074,
+      "step": 1517
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963349195217306,
+      "loss": 0.7085,
+      "step": 1518
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019633003845499848,
+      "loss": 0.7408,
+      "step": 1519
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001963251542009402,
+      "loss": 0.7558,
+      "step": 1520
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019632026675971736,
+      "loss": 0.6885,
+      "step": 1521
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019631537613149161,
+      "loss": 0.7965,
+      "step": 1522
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019631048231642489,
+      "loss": 0.7393,
+      "step": 1523
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019630558531467905,
+      "loss": 0.7729,
+      "step": 1524
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019630068512641608,
+      "loss": 0.7613,
+      "step": 1525
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001962957817517982,
+      "loss": 0.7609,
+      "step": 1526
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001962908751909876,
+      "loss": 0.7777,
+      "step": 1527
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019628596544414665,
+      "loss": 0.7714,
+      "step": 1528
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001962810525114378,
+      "loss": 0.7651,
+      "step": 1529
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001962761363930236,
+      "loss": 0.777,
+      "step": 1530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019627121708906667,
+      "loss": 0.7268,
+      "step": 1531
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019626629459972987,
+      "loss": 0.7542,
+      "step": 1532
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019626136892517598,
+      "loss": 0.8172,
+      "step": 1533
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019625644006556806,
+      "loss": 0.731,
+      "step": 1534
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019625150802106913,
+      "loss": 0.7691,
+      "step": 1535
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019624657279184243,
+      "loss": 0.8145,
+      "step": 1536
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001962416343780512,
+      "loss": 0.7575,
+      "step": 1537
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019623669277985887,
+      "loss": 0.7934,
+      "step": 1538
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019623174799742894,
+      "loss": 0.8042,
+      "step": 1539
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019622680003092503,
+      "loss": 0.7596,
+      "step": 1540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019622184888051086,
+      "loss": 0.7777,
+      "step": 1541
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019621689454635025,
+      "loss": 0.8296,
+      "step": 1542
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001962119370286071,
+      "loss": 0.7923,
+      "step": 1543
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001962069763274455,
+      "loss": 0.7504,
+      "step": 1544
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019620201244302952,
+      "loss": 0.6759,
+      "step": 1545
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019619704537552344,
+      "loss": 0.7424,
+      "step": 1546
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001961920751250916,
+      "loss": 0.7481,
+      "step": 1547
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019618710169189843,
+      "loss": 0.7799,
+      "step": 1548
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019618212507610853,
+      "loss": 0.7389,
+      "step": 1549
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019617714527788652,
+      "loss": 0.7473,
+      "step": 1550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019617216229739722,
+      "loss": 0.7767,
+      "step": 1551
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019616717613480552,
+      "loss": 0.8095,
+      "step": 1552
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001961621867902763,
+      "loss": 0.8234,
+      "step": 1553
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019615719426397475,
+      "loss": 0.7389,
+      "step": 1554
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019615219855606596,
+      "loss": 0.7472,
+      "step": 1555
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019614719966671532,
+      "loss": 0.775,
+      "step": 1556
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001961421975960882,
+      "loss": 0.793,
+      "step": 1557
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019613719234435012,
+      "loss": 0.7804,
+      "step": 1558
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019613218391166664,
+      "loss": 0.8263,
+      "step": 1559
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019612717229820347,
+      "loss": 0.7598,
+      "step": 1560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001961221575041265,
+      "loss": 0.7989,
+      "step": 1561
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019611713952960168,
+      "loss": 0.7903,
+      "step": 1562
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019611211837479493,
+      "loss": 0.7988,
+      "step": 1563
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019610709403987246,
+      "loss": 0.751,
+      "step": 1564
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001961020665250005,
+      "loss": 0.7984,
+      "step": 1565
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019609703583034542,
+      "loss": 0.902,
+      "step": 1566
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019609200195607363,
+      "loss": 0.7783,
+      "step": 1567
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001960869649023517,
+      "loss": 0.7352,
+      "step": 1568
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019608192466934635,
+      "loss": 0.7092,
+      "step": 1569
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019607688125722426,
+      "loss": 0.7046,
+      "step": 1570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019607183466615236,
+      "loss": 0.7542,
+      "step": 1571
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019606678489629762,
+      "loss": 0.8297,
+      "step": 1572
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019606173194782712,
+      "loss": 0.7637,
+      "step": 1573
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019605667582090806,
+      "loss": 0.7838,
+      "step": 1574
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019605161651570772,
+      "loss": 0.8752,
+      "step": 1575
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019604655403239352,
+      "loss": 0.7874,
+      "step": 1576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001960414883711329,
+      "loss": 0.7794,
+      "step": 1577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019603641953209357,
+      "loss": 0.8109,
+      "step": 1578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001960313475154432,
+      "loss": 0.7187,
+      "step": 1579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001960262723213496,
+      "loss": 0.7824,
+      "step": 1580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001960211939499807,
+      "loss": 0.7749,
+      "step": 1581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019601611240150453,
+      "loss": 0.7483,
+      "step": 1582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019601102767608923,
+      "loss": 0.8099,
+      "step": 1583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019600593977390306,
+      "loss": 0.8243,
+      "step": 1584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019600084869511434,
+      "loss": 0.8711,
+      "step": 1585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019599575443989152,
+      "loss": 0.827,
+      "step": 1586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019599065700840317,
+      "loss": 0.7581,
+      "step": 1587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000195985556400818,
+      "loss": 0.7821,
+      "step": 1588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019598045261730467,
+      "loss": 0.7281,
+      "step": 1589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001959753456580321,
+      "loss": 0.7767,
+      "step": 1590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019597023552316928,
+      "loss": 0.7615,
+      "step": 1591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019596512221288533,
+      "loss": 0.8057,
+      "step": 1592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019596000572734933,
+      "loss": 0.6927,
+      "step": 1593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019595488606673068,
+      "loss": 0.7775,
+      "step": 1594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001959497632311987,
+      "loss": 0.8287,
+      "step": 1595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019594463722092293,
+      "loss": 0.7601,
+      "step": 1596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000195939508036073,
+      "loss": 0.8217,
+      "step": 1597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019593437567681855,
+      "loss": 0.7744,
+      "step": 1598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019592924014332947,
+      "loss": 0.7368,
+      "step": 1599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019592410143577565,
+      "loss": 0.7555,
+      "step": 1600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001959189595543271,
+      "loss": 0.7527,
+      "step": 1601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019591381449915397,
+      "loss": 0.802,
+      "step": 1602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019590866627042653,
+      "loss": 0.8041,
+      "step": 1603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001959035148683151,
+      "loss": 0.746,
+      "step": 1604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019589836029299006,
+      "loss": 0.8344,
+      "step": 1605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019589320254462207,
+      "loss": 0.7961,
+      "step": 1606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019588804162338174,
+      "loss": 0.7532,
+      "step": 1607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019588287752943983,
+      "loss": 0.7953,
+      "step": 1608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019587771026296716,
+      "loss": 0.8054,
+      "step": 1609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019587253982413479,
+      "loss": 0.733,
+      "step": 1610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019586736621311378,
+      "loss": 0.7861,
+      "step": 1611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019586218943007524,
+      "loss": 0.7792,
+      "step": 1612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019585700947519054,
+      "loss": 0.8229,
+      "step": 1613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019585182634863102,
+      "loss": 0.7914,
+      "step": 1614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019584664005056822,
+      "loss": 0.7783,
+      "step": 1615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001958414505811737,
+      "loss": 0.7863,
+      "step": 1616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019583625794061918,
+      "loss": 0.7399,
+      "step": 1617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001958310621290765,
+      "loss": 0.7926,
+      "step": 1618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019582586314671754,
+      "loss": 0.8399,
+      "step": 1619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019582066099371433,
+      "loss": 0.7179,
+      "step": 1620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000195815455670239,
+      "loss": 0.7279,
+      "step": 1621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001958102471764638,
+      "loss": 0.7404,
+      "step": 1622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019580503551256103,
+      "loss": 0.7521,
+      "step": 1623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019579982067870316,
+      "loss": 0.7334,
+      "step": 1624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019579460267506274,
+      "loss": 0.7648,
+      "step": 1625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001957893815018124,
+      "loss": 0.7046,
+      "step": 1626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019578415715912491,
+      "loss": 0.785,
+      "step": 1627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019577892964717312,
+      "loss": 0.7026,
+      "step": 1628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019577369896612998,
+      "loss": 0.729,
+      "step": 1629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001957684651161686,
+      "loss": 0.7562,
+      "step": 1630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019576322809746214,
+      "loss": 0.8113,
+      "step": 1631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001957579879101839,
+      "loss": 0.8194,
+      "step": 1632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019575274455450718,
+      "loss": 0.7517,
+      "step": 1633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019574749803060557,
+      "loss": 0.7662,
+      "step": 1634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019574224833865262,
+      "loss": 0.7564,
+      "step": 1635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019573699547882205,
+      "loss": 0.7371,
+      "step": 1636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019573173945128765,
+      "loss": 0.7225,
+      "step": 1637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001957264802562233,
+      "loss": 0.7889,
+      "step": 1638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019572121789380304,
+      "loss": 0.7763,
+      "step": 1639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019571595236420102,
+      "loss": 0.7841,
+      "step": 1640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019571068366759143,
+      "loss": 0.7431,
+      "step": 1641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001957054118041486,
+      "loss": 0.7655,
+      "step": 1642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019570013677404696,
+      "loss": 0.8047,
+      "step": 1643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019569485857746105,
+      "loss": 0.7545,
+      "step": 1644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019568957721456554,
+      "loss": 0.7608,
+      "step": 1645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019568429268553513,
+      "loss": 0.7028,
+      "step": 1646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019567900499054472,
+      "loss": 0.7774,
+      "step": 1647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019567371412976923,
+      "loss": 0.7023,
+      "step": 1648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019566842010338372,
+      "loss": 0.7728,
+      "step": 1649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019566312291156341,
+      "loss": 0.7937,
+      "step": 1650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019565782255448352,
+      "loss": 0.8152,
+      "step": 1651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019565251903231942,
+      "loss": 0.7915,
+      "step": 1652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019564721234524663,
+      "loss": 0.8577,
+      "step": 1653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001956419024934407,
+      "loss": 0.7754,
+      "step": 1654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019563658947707735,
+      "loss": 0.709,
+      "step": 1655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019563127329633235,
+      "loss": 0.751,
+      "step": 1656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019562595395138162,
+      "loss": 0.7174,
+      "step": 1657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019562063144240113,
+      "loss": 0.8105,
+      "step": 1658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019561530576956703,
+      "loss": 0.672,
+      "step": 1659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019560997693305552,
+      "loss": 0.8472,
+      "step": 1660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019560464493304292,
+      "loss": 0.7774,
+      "step": 1661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019559930976970562,
+      "loss": 0.7144,
+      "step": 1662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019559397144322017,
+      "loss": 0.756,
+      "step": 1663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019558862995376326,
+      "loss": 0.7105,
+      "step": 1664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019558328530151152,
+      "loss": 0.7397,
+      "step": 1665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019557793748664187,
+      "loss": 0.8143,
+      "step": 1666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019557258650933123,
+      "loss": 0.7539,
+      "step": 1667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019556723236975664,
+      "loss": 0.7197,
+      "step": 1668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019556187506809526,
+      "loss": 0.7958,
+      "step": 1669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019555651460452438,
+      "loss": 0.9077,
+      "step": 1670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019555115097922134,
+      "loss": 0.7741,
+      "step": 1671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001955457841923636,
+      "loss": 0.8122,
+      "step": 1672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019554041424412873,
+      "loss": 0.7628,
+      "step": 1673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019553504113469446,
+      "loss": 0.7435,
+      "step": 1674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019552966486423854,
+      "loss": 0.8133,
+      "step": 1675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019552428543293882,
+      "loss": 0.785,
+      "step": 1676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019551890284097334,
+      "loss": 0.7094,
+      "step": 1677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001955135170885202,
+      "loss": 0.7229,
+      "step": 1678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019550812817575755,
+      "loss": 0.8139,
+      "step": 1679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019550273610286376,
+      "loss": 0.7605,
+      "step": 1680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001954973408700172,
+      "loss": 0.7961,
+      "step": 1681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001954919424773964,
+      "loss": 0.7558,
+      "step": 1682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019548654092518,
+      "loss": 0.7816,
+      "step": 1683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019548113621354666,
+      "loss": 0.7855,
+      "step": 1684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019547572834267526,
+      "loss": 0.803,
+      "step": 1685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019547031731274473,
+      "loss": 0.7996,
+      "step": 1686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001954649031239341,
+      "loss": 0.8445,
+      "step": 1687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019545948577642252,
+      "loss": 0.7658,
+      "step": 1688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019545406527038923,
+      "loss": 0.7842,
+      "step": 1689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019544864160601357,
+      "loss": 0.7812,
+      "step": 1690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000195443214783475,
+      "loss": 0.7294,
+      "step": 1691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001954377848029531,
+      "loss": 0.7576,
+      "step": 1692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019543235166462753,
+      "loss": 0.7897,
+      "step": 1693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019542691536867803,
+      "loss": 0.7448,
+      "step": 1694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019542147591528451,
+      "loss": 0.8915,
+      "step": 1695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019541603330462692,
+      "loss": 0.7487,
+      "step": 1696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019541058753688538,
+      "loss": 0.7334,
+      "step": 1697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019540513861224004,
+      "loss": 0.7862,
+      "step": 1698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001953996865308712,
+      "loss": 0.7786,
+      "step": 1699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019539423129295925,
+      "loss": 0.8008,
+      "step": 1700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019538877289868473,
+      "loss": 0.6999,
+      "step": 1701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019538331134822821,
+      "loss": 0.6779,
+      "step": 1702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001953778466417704,
+      "loss": 0.7289,
+      "step": 1703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001953723787794921,
+      "loss": 0.7585,
+      "step": 1704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019536690776157428,
+      "loss": 0.7273,
+      "step": 1705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001953614335881979,
+      "loss": 0.7949,
+      "step": 1706
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019535595625954413,
+      "loss": 0.8446,
+      "step": 1707
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001953504757757942,
+      "loss": 0.7728,
+      "step": 1708
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019534499213712942,
+      "loss": 0.7796,
+      "step": 1709
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019533950534373124,
+      "loss": 0.7596,
+      "step": 1710
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019533401539578122,
+      "loss": 0.7714,
+      "step": 1711
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019532852229346101,
+      "loss": 0.7595,
+      "step": 1712
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019532302603695234,
+      "loss": 0.7641,
+      "step": 1713
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001953175266264371,
+      "loss": 0.8068,
+      "step": 1714
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019531202406209718,
+      "loss": 0.8001,
+      "step": 1715
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019530651834411474,
+      "loss": 0.8243,
+      "step": 1716
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001953010094726719,
+      "loss": 0.8574,
+      "step": 1717
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019529549744795097,
+      "loss": 0.7073,
+      "step": 1718
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019528998227013426,
+      "loss": 0.8127,
+      "step": 1719
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001952844639394043,
+      "loss": 0.819,
+      "step": 1720
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019527894245594373,
+      "loss": 0.7172,
+      "step": 1721
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019527341781993513,
+      "loss": 0.755,
+      "step": 1722
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001952678900315614,
+      "loss": 0.886,
+      "step": 1723
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019526235909100533,
+      "loss": 0.7199,
+      "step": 1724
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019525682499845005,
+      "loss": 0.7603,
+      "step": 1725
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019525128775407859,
+      "loss": 0.7335,
+      "step": 1726
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019524574735807417,
+      "loss": 0.8546,
+      "step": 1727
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019524020381062016,
+      "loss": 0.7944,
+      "step": 1728
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001952346571118999,
+      "loss": 0.7953,
+      "step": 1729
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000195229107262097,
+      "loss": 0.7037,
+      "step": 1730
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019522355426139504,
+      "loss": 0.7717,
+      "step": 1731
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019521799810997775,
+      "loss": 0.741,
+      "step": 1732
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019521243880802903,
+      "loss": 0.8449,
+      "step": 1733
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019520687635573273,
+      "loss": 0.7784,
+      "step": 1734
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019520131075327298,
+      "loss": 0.7183,
+      "step": 1735
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001951957420008339,
+      "loss": 0.7054,
+      "step": 1736
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019519017009859976,
+      "loss": 0.7845,
+      "step": 1737
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001951845950467549,
+      "loss": 0.7316,
+      "step": 1738
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001951790168454838,
+      "loss": 0.7709,
+      "step": 1739
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019517343549497101,
+      "loss": 0.6963,
+      "step": 1740
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019516785099540123,
+      "loss": 0.7248,
+      "step": 1741
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001951622633469592,
+      "loss": 0.7471,
+      "step": 1742
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019515667254982983,
+      "loss": 0.7915,
+      "step": 1743
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019515107860419812,
+      "loss": 0.8656,
+      "step": 1744
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019514548151024913,
+      "loss": 0.7498,
+      "step": 1745
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019513988126816805,
+      "loss": 0.8191,
+      "step": 1746
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001951342778781402,
+      "loss": 0.7255,
+      "step": 1747
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000195128671340351,
+      "loss": 0.797,
+      "step": 1748
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001951230616549859,
+      "loss": 0.7883,
+      "step": 1749
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019511744882223057,
+      "loss": 0.7808,
+      "step": 1750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019511183284227072,
+      "loss": 0.8009,
+      "step": 1751
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019510621371529208,
+      "loss": 0.7986,
+      "step": 1752
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001951005914414807,
+      "loss": 0.7581,
+      "step": 1753
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019509496602102252,
+      "loss": 0.7189,
+      "step": 1754
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950893374541037,
+      "loss": 0.7676,
+      "step": 1755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950837057409105,
+      "loss": 0.7957,
+      "step": 1756
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950780708816292,
+      "loss": 0.8188,
+      "step": 1757
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950724328764463,
+      "loss": 0.773,
+      "step": 1758
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019506679172554833,
+      "loss": 0.7583,
+      "step": 1759
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019506114742912192,
+      "loss": 0.7478,
+      "step": 1760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019505549998735387,
+      "loss": 0.758,
+      "step": 1761
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000195049849400431,
+      "loss": 0.783,
+      "step": 1762
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950441956685403,
+      "loss": 0.7958,
+      "step": 1763
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950385387918688,
+      "loss": 0.7608,
+      "step": 1764
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019503287877060371,
+      "loss": 0.7565,
+      "step": 1765
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019502721560493233,
+      "loss": 0.7286,
+      "step": 1766
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019502154929504196,
+      "loss": 0.7857,
+      "step": 1767
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019501587984112014,
+      "loss": 0.8058,
+      "step": 1768
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950102072433545,
+      "loss": 0.7651,
+      "step": 1769
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950045315019326,
+      "loss": 0.7046,
+      "step": 1770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019499885261704236,
+      "loss": 0.8094,
+      "step": 1771
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019499317058887164,
+      "loss": 0.789,
+      "step": 1772
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019498748541760846,
+      "loss": 0.7328,
+      "step": 1773
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019498179710344087,
+      "loss": 0.7687,
+      "step": 1774
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019497610564655715,
+      "loss": 0.76,
+      "step": 1775
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019497041104714558,
+      "loss": 0.7873,
+      "step": 1776
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001949647133053946,
+      "loss": 0.7961,
+      "step": 1777
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001949590124214927,
+      "loss": 0.736,
+      "step": 1778
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019495330839562852,
+      "loss": 0.732,
+      "step": 1779
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019494760122799084,
+      "loss": 0.6641,
+      "step": 1780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019494189091876846,
+      "loss": 0.7347,
+      "step": 1781
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019493617746815033,
+      "loss": 0.7522,
+      "step": 1782
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019493046087632547,
+      "loss": 0.7122,
+      "step": 1783
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019492474114348306,
+      "loss": 0.7657,
+      "step": 1784
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001949190182698123,
+      "loss": 0.7532,
+      "step": 1785
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019491329225550263,
+      "loss": 0.7689,
+      "step": 1786
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001949075631007434,
+      "loss": 0.7496,
+      "step": 1787
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001949018308057243,
+      "loss": 0.7484,
+      "step": 1788
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001948960953706349,
+      "loss": 0.8441,
+      "step": 1789
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019489035679566503,
+      "loss": 0.7565,
+      "step": 1790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001948846150810045,
+      "loss": 0.7346,
+      "step": 1791
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019487887022684336,
+      "loss": 0.7743,
+      "step": 1792
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019487312223337165,
+      "loss": 0.809,
+      "step": 1793
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019486737110077957,
+      "loss": 0.6915,
+      "step": 1794
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001948616168292574,
+      "loss": 0.7398,
+      "step": 1795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019485585941899557,
+      "loss": 0.7116,
+      "step": 1796
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019485009887018455,
+      "loss": 0.7715,
+      "step": 1797
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019484433518301495,
+      "loss": 0.682,
+      "step": 1798
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019483856835767746,
+      "loss": 0.7979,
+      "step": 1799
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019483279839436292,
+      "loss": 0.7332,
+      "step": 1800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001948270252932622,
+      "loss": 0.7869,
+      "step": 1801
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019482124905456636,
+      "loss": 0.7338,
+      "step": 1802
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019481546967846654,
+      "loss": 0.7493,
+      "step": 1803
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019480968716515388,
+      "loss": 0.7844,
+      "step": 1804
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019480390151481978,
+      "loss": 0.8204,
+      "step": 1805
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019479811272765567,
+      "loss": 0.7114,
+      "step": 1806
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019479232080385305,
+      "loss": 0.7618,
+      "step": 1807
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001947865257436036,
+      "loss": 0.7874,
+      "step": 1808
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000194780727547099,
+      "loss": 0.7767,
+      "step": 1809
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019477492621453122,
+      "loss": 0.7185,
+      "step": 1810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001947691217460921,
+      "loss": 0.8103,
+      "step": 1811
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019476331414197372,
+      "loss": 0.7964,
+      "step": 1812
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019475750340236826,
+      "loss": 0.806,
+      "step": 1813
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000194751689527468,
+      "loss": 0.7937,
+      "step": 1814
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019474587251746525,
+      "loss": 0.776,
+      "step": 1815
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019474005237255253,
+      "loss": 0.7791,
+      "step": 1816
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001947342290929224,
+      "loss": 0.7471,
+      "step": 1817
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019472840267876752,
+      "loss": 0.7158,
+      "step": 1818
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001947225731302807,
+      "loss": 0.7588,
+      "step": 1819
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019471674044765482,
+      "loss": 0.7377,
+      "step": 1820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019471090463108284,
+      "loss": 0.7786,
+      "step": 1821
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001947050656807579,
+      "loss": 0.7258,
+      "step": 1822
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019469922359687317,
+      "loss": 0.8091,
+      "step": 1823
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019469337837962192,
+      "loss": 0.7539,
+      "step": 1824
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019468753002919762,
+      "loss": 0.7894,
+      "step": 1825
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001946816785457937,
+      "loss": 0.7702,
+      "step": 1826
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019467582392960385,
+      "loss": 0.7889,
+      "step": 1827
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019466996618082175,
+      "loss": 0.8297,
+      "step": 1828
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019466410529964123,
+      "loss": 0.7386,
+      "step": 1829
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019465824128625617,
+      "loss": 0.7576,
+      "step": 1830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019465237414086067,
+      "loss": 0.7027,
+      "step": 1831
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019464650386364878,
+      "loss": 0.722,
+      "step": 1832
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019464063045481478,
+      "loss": 0.7718,
+      "step": 1833
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000194634753914553,
+      "loss": 0.7246,
+      "step": 1834
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019462887424305785,
+      "loss": 0.7818,
+      "step": 1835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019462299144052392,
+      "loss": 0.7216,
+      "step": 1836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019461710550714584,
+      "loss": 0.6824,
+      "step": 1837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019461121644311836,
+      "loss": 0.7187,
+      "step": 1838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001946053242486363,
+      "loss": 0.7697,
+      "step": 1839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019459942892389472,
+      "loss": 0.7455,
+      "step": 1840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019459353046908856,
+      "loss": 0.7832,
+      "step": 1841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019458762888441308,
+      "loss": 0.8186,
+      "step": 1842
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019458172417006347,
+      "loss": 0.7325,
+      "step": 1843
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019457581632623518,
+      "loss": 0.7603,
+      "step": 1844
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019456990535312362,
+      "loss": 0.7564,
+      "step": 1845
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019456399125092441,
+      "loss": 0.8493,
+      "step": 1846
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001945580740198332,
+      "loss": 0.7713,
+      "step": 1847
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019455215366004583,
+      "loss": 0.7712,
+      "step": 1848
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019454623017175812,
+      "loss": 0.7411,
+      "step": 1849
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019454030355516616,
+      "loss": 0.8119,
+      "step": 1850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019453437381046593,
+      "loss": 0.7821,
+      "step": 1851
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019452844093785369,
+      "loss": 0.7662,
+      "step": 1852
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001945225049375258,
+      "loss": 0.7594,
+      "step": 1853
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019451656580967856,
+      "loss": 0.8347,
+      "step": 1854
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019451062355450854,
+      "loss": 0.7688,
+      "step": 1855
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019450467817221236,
+      "loss": 0.7225,
+      "step": 1856
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019449872966298672,
+      "loss": 0.7406,
+      "step": 1857
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019449277802702848,
+      "loss": 0.7524,
+      "step": 1858
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001944868232645345,
+      "loss": 0.7591,
+      "step": 1859
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019448086537570184,
+      "loss": 0.7908,
+      "step": 1860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019447490436072764,
+      "loss": 0.7549,
+      "step": 1861
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019446894021980912,
+      "loss": 0.7645,
+      "step": 1862
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019446297295314365,
+      "loss": 0.7007,
+      "step": 1863
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001944570025609286,
+      "loss": 0.801,
+      "step": 1864
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001944510290433616,
+      "loss": 0.7754,
+      "step": 1865
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019444505240064032,
+      "loss": 0.7865,
+      "step": 1866
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019443907263296236,
+      "loss": 0.7202,
+      "step": 1867
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001944330897405257,
+      "loss": 0.8356,
+      "step": 1868
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019442710372352833,
+      "loss": 0.7117,
+      "step": 1869
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001944211145821682,
+      "loss": 0.6704,
+      "step": 1870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019441512231664355,
+      "loss": 0.7469,
+      "step": 1871
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019440912692715264,
+      "loss": 0.74,
+      "step": 1872
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001944031284138938,
+      "loss": 0.7533,
+      "step": 1873
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001943971267770656,
+      "loss": 0.7733,
+      "step": 1874
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001943911220168665,
+      "loss": 0.7892,
+      "step": 1875
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019438511413349526,
+      "loss": 0.776,
+      "step": 1876
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019437910312715065,
+      "loss": 0.769,
+      "step": 1877
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019437308899803154,
+      "loss": 0.7193,
+      "step": 1878
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00019436707174633696,
+      "loss": 0.8029,
+      "step": 1879
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019436105137226602,
+      "loss": 0.7893,
+      "step": 1880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019435502787601782,
+      "loss": 0.7411,
+      "step": 1881
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019434900125779177,
+      "loss": 0.7993,
+      "step": 1882
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019434297151778724,
+      "loss": 0.7161,
+      "step": 1883
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019433693865620374,
+      "loss": 0.7569,
+      "step": 1884
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019433090267324086,
+      "loss": 0.7453,
+      "step": 1885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019432486356909832,
+      "loss": 0.781,
+      "step": 1886
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019431882134397598,
+      "loss": 0.7439,
+      "step": 1887
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019431277599807372,
+      "loss": 0.7805,
+      "step": 1888
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019430672753159155,
+      "loss": 0.8329,
+      "step": 1889
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001943006759447297,
+      "loss": 0.7688,
+      "step": 1890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001942946212376883,
+      "loss": 0.7473,
+      "step": 1891
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001942885634106677,
+      "loss": 0.764,
+      "step": 1892
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019428250246386836,
+      "loss": 0.7311,
+      "step": 1893
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019427643839749084,
+      "loss": 0.7322,
+      "step": 1894
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019427037121173574,
+      "loss": 0.688,
+      "step": 1895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019426430090680388,
+      "loss": 0.7618,
+      "step": 1896
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019425822748289601,
+      "loss": 0.8215,
+      "step": 1897
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019425215094021316,
+      "loss": 0.7758,
+      "step": 1898
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019424607127895637,
+      "loss": 0.7278,
+      "step": 1899
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001942399884993268,
+      "loss": 0.7715,
+      "step": 1900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001942339026015257,
+      "loss": 0.774,
+      "step": 1901
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019422781358575443,
+      "loss": 0.7269,
+      "step": 1902
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019422172145221452,
+      "loss": 0.7616,
+      "step": 1903
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019421562620110748,
+      "loss": 0.752,
+      "step": 1904
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019420952783263503,
+      "loss": 0.8105,
+      "step": 1905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001942034263469989,
+      "loss": 0.7085,
+      "step": 1906
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019419732174440104,
+      "loss": 0.8261,
+      "step": 1907
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001941912140250434,
+      "loss": 0.7672,
+      "step": 1908
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019418510318912807,
+      "loss": 0.7537,
+      "step": 1909
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001941789892368572,
+      "loss": 0.8291,
+      "step": 1910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019417287216843317,
+      "loss": 0.8126,
+      "step": 1911
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019416675198405834,
+      "loss": 0.754,
+      "step": 1912
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019416062868393518,
+      "loss": 0.7134,
+      "step": 1913
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019415450226826635,
+      "loss": 0.7043,
+      "step": 1914
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019414837273725453,
+      "loss": 0.7559,
+      "step": 1915
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001941422400911025,
+      "loss": 0.7155,
+      "step": 1916
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019413610433001327,
+      "loss": 0.7174,
+      "step": 1917
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019412996545418976,
+      "loss": 0.6592,
+      "step": 1918
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019412382346383514,
+      "loss": 0.8389,
+      "step": 1919
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001941176783591526,
+      "loss": 0.768,
+      "step": 1920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001941115301403455,
+      "loss": 0.7568,
+      "step": 1921
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00019410537880761724,
+      "loss": 0.7781,
+      "step": 1922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001940992243611714,
+      "loss": 0.7672,
+      "step": 1923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019409306680121156,
+      "loss": 0.7251,
+      "step": 1924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019408690612794148,
+      "loss": 0.8235,
+      "step": 1925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019408074234156502,
+      "loss": 0.7173,
+      "step": 1926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019407457544228608,
+      "loss": 0.7873,
+      "step": 1927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019406840543030878,
+      "loss": 0.785,
+      "step": 1928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001940622323058372,
+      "loss": 0.8044,
+      "step": 1929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001940560560690756,
+      "loss": 0.7276,
+      "step": 1930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019404987672022844,
+      "loss": 0.7877,
+      "step": 1931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019404369425950004,
+      "loss": 0.7743,
+      "step": 1932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000194037508687095,
+      "loss": 0.7035,
+      "step": 1933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019403132000321804,
+      "loss": 0.6944,
+      "step": 1934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001940251282080739,
+      "loss": 0.7359,
+      "step": 1935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019401893330186741,
+      "loss": 0.8501,
+      "step": 1936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019401273528480358,
+      "loss": 0.6684,
+      "step": 1937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019400653415708752,
+      "loss": 0.7933,
+      "step": 1938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019400032991892437,
+      "loss": 0.7854,
+      "step": 1939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019399412257051942,
+      "loss": 0.7334,
+      "step": 1940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019398791211207804,
+      "loss": 0.7497,
+      "step": 1941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019398169854380575,
+      "loss": 0.7515,
+      "step": 1942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019397548186590814,
+      "loss": 0.837,
+      "step": 1943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019396926207859084,
+      "loss": 0.7454,
+      "step": 1944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019396303918205974,
+      "loss": 0.7085,
+      "step": 1945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019395681317652068,
+      "loss": 0.7766,
+      "step": 1946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019395058406217972,
+      "loss": 0.7443,
+      "step": 1947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001939443518392429,
+      "loss": 0.8331,
+      "step": 1948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019393811650791646,
+      "loss": 0.7972,
+      "step": 1949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019393187806840672,
+      "loss": 0.7421,
+      "step": 1950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019392563652092005,
+      "loss": 0.7552,
+      "step": 1951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019391939186566304,
+      "loss": 0.7762,
+      "step": 1952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001939131441028423,
+      "loss": 0.7062,
+      "step": 1953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019390689323266447,
+      "loss": 0.7616,
+      "step": 1954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019390063925533646,
+      "loss": 0.7368,
+      "step": 1955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019389438217106518,
+      "loss": 0.8178,
+      "step": 1956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019388812198005764,
+      "loss": 0.8426,
+      "step": 1957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019388185868252097,
+      "loss": 0.6997,
+      "step": 1958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019387559227866246,
+      "loss": 0.7197,
+      "step": 1959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019386932276868938,
+      "loss": 0.7188,
+      "step": 1960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019386305015280924,
+      "loss": 0.7916,
+      "step": 1961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019385677443122955,
+      "loss": 0.7566,
+      "step": 1962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019385049560415794,
+      "loss": 0.7596,
+      "step": 1963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019384421367180216,
+      "loss": 0.7263,
+      "step": 1964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00019383792863437015,
+      "loss": 0.7871,
+      "step": 1965
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019383164049206975,
+      "loss": 0.7495,
+      "step": 1966
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001938253492451091,
+      "loss": 0.7421,
+      "step": 1967
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019381905489369633,
+      "loss": 0.6849,
+      "step": 1968
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001938127574380397,
+      "loss": 0.7664,
+      "step": 1969
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001938064568783476,
+      "loss": 0.7494,
+      "step": 1970
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019380015321482846,
+      "loss": 0.73,
+      "step": 1971
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019379384644769093,
+      "loss": 0.766,
+      "step": 1972
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001937875365771436,
+      "loss": 0.7808,
+      "step": 1973
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001937812236033953,
+      "loss": 0.7946,
+      "step": 1974
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019377490752665486,
+      "loss": 0.742,
+      "step": 1975
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019376858834713133,
+      "loss": 0.7873,
+      "step": 1976
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019376226606503374,
+      "loss": 0.8364,
+      "step": 1977
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019375594068057132,
+      "loss": 0.7833,
+      "step": 1978
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019374961219395336,
+      "loss": 0.7525,
+      "step": 1979
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001937432806053892,
+      "loss": 0.8181,
+      "step": 1980
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001937369459150884,
+      "loss": 0.7175,
+      "step": 1981
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019373060812326052,
+      "loss": 0.7171,
+      "step": 1982
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001937242672301153,
+      "loss": 0.8029,
+      "step": 1983
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001937179232358625,
+      "loss": 0.688,
+      "step": 1984
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019371157614071207,
+      "loss": 0.7855,
+      "step": 1985
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000193705225944874,
+      "loss": 0.7559,
+      "step": 1986
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019369887264855838,
+      "loss": 0.763,
+      "step": 1987
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019369251625197547,
+      "loss": 0.789,
+      "step": 1988
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019368615675533553,
+      "loss": 0.7814,
+      "step": 1989
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019367979415884902,
+      "loss": 0.7426,
+      "step": 1990
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019367342846272648,
+      "loss": 0.7541,
+      "step": 1991
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019366705966717852,
+      "loss": 0.7699,
+      "step": 1992
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019366068777241583,
+      "loss": 0.8155,
+      "step": 1993
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001936543127786493,
+      "loss": 0.7707,
+      "step": 1994
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001936479346860898,
+      "loss": 0.7539,
+      "step": 1995
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019364155349494843,
+      "loss": 0.7001,
+      "step": 1996
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019363516920543628,
+      "loss": 0.7233,
+      "step": 1997
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019362878181776463,
+      "loss": 0.7713,
+      "step": 1998
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001936223913321448,
+      "loss": 0.8308,
+      "step": 1999
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019361599774878822,
+      "loss": 0.7633,
+      "step": 2000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019360960106790643,
+      "loss": 0.7324,
+      "step": 2001
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019360320128971117,
+      "loss": 0.7036,
+      "step": 2002
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001935967984144141,
+      "loss": 0.7737,
+      "step": 2003
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001935903924422271,
+      "loss": 0.7084,
+      "step": 2004
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019358398337336212,
+      "loss": 0.7164,
+      "step": 2005
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019357757120803124,
+      "loss": 0.8324,
+      "step": 2006
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019357115594644664,
+      "loss": 0.7957,
+      "step": 2007
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00019356473758882052,
+      "loss": 0.743,
+      "step": 2008
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019355831613536531,
+      "loss": 0.7896,
+      "step": 2009
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019355189158629346,
+      "loss": 0.7239,
+      "step": 2010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019354546394181754,
+      "loss": 0.7183,
+      "step": 2011
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019353903320215024,
+      "loss": 0.7735,
+      "step": 2012
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001935325993675043,
+      "loss": 0.7668,
+      "step": 2013
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019352616243809262,
+      "loss": 0.6915,
+      "step": 2014
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001935197224141282,
+      "loss": 0.7451,
+      "step": 2015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019351327929582413,
+      "loss": 0.706,
+      "step": 2016
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019350683308339356,
+      "loss": 0.6836,
+      "step": 2017
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019350038377704979,
+      "loss": 0.7895,
+      "step": 2018
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019349393137700624,
+      "loss": 0.8115,
+      "step": 2019
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019348747588347637,
+      "loss": 0.7315,
+      "step": 2020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019348101729667382,
+      "loss": 0.7418,
+      "step": 2021
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019347455561681223,
+      "loss": 0.6854,
+      "step": 2022
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019346809084410543,
+      "loss": 0.7243,
+      "step": 2023
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001934616229787673,
+      "loss": 0.7819,
+      "step": 2024
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001934551520210119,
+      "loss": 0.7386,
+      "step": 2025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019344867797105333,
+      "loss": 0.8092,
+      "step": 2026
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019344220082910573,
+      "loss": 0.7271,
+      "step": 2027
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001934357205953835,
+      "loss": 0.7233,
+      "step": 2028
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019342923727010098,
+      "loss": 0.7853,
+      "step": 2029
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019342275085347277,
+      "loss": 0.692,
+      "step": 2030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019341626134571343,
+      "loss": 0.7935,
+      "step": 2031
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019340976874703767,
+      "loss": 0.7261,
+      "step": 2032
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019340327305766036,
+      "loss": 0.7036,
+      "step": 2033
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001933967742777964,
+      "loss": 0.8032,
+      "step": 2034
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019339027240766084,
+      "loss": 0.7902,
+      "step": 2035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001933837674474688,
+      "loss": 0.7534,
+      "step": 2036
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019337725939743545,
+      "loss": 0.688,
+      "step": 2037
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019337074825777624,
+      "loss": 0.7907,
+      "step": 2038
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019336423402870653,
+      "loss": 0.7551,
+      "step": 2039
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001933577167104419,
+      "loss": 0.7951,
+      "step": 2040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019335119630319798,
+      "loss": 0.8153,
+      "step": 2041
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019334467280719052,
+      "loss": 0.7308,
+      "step": 2042
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001933381462226353,
+      "loss": 0.7116,
+      "step": 2043
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019333161654974836,
+      "loss": 0.7594,
+      "step": 2044
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001933250837887457,
+      "loss": 0.7979,
+      "step": 2045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019331854793984353,
+      "loss": 0.671,
+      "step": 2046
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019331200900325803,
+      "loss": 0.7667,
+      "step": 2047
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001933054669792056,
+      "loss": 0.6835,
+      "step": 2048
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001932989218679027,
+      "loss": 0.7776,
+      "step": 2049
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00019329237366956588,
+      "loss": 0.7975,
+      "step": 2050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001932858223844118,
+      "loss": 0.7134,
+      "step": 2051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019327926801265724,
+      "loss": 0.7856,
+      "step": 2052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019327271055451906,
+      "loss": 0.763,
+      "step": 2053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019326615001021423,
+      "loss": 0.7528,
+      "step": 2054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019325958637995982,
+      "loss": 0.7262,
+      "step": 2055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000193253019663973,
+      "loss": 0.759,
+      "step": 2056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019324644986247106,
+      "loss": 0.7985,
+      "step": 2057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001932398769756714,
+      "loss": 0.7587,
+      "step": 2058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019323330100379146,
+      "loss": 0.8345,
+      "step": 2059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001932267219470488,
+      "loss": 0.7535,
+      "step": 2060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019322013980566119,
+      "loss": 0.7978,
+      "step": 2061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019321355457984637,
+      "loss": 0.7847,
+      "step": 2062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001932069662698222,
+      "loss": 0.673,
+      "step": 2063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001932003748758067,
+      "loss": 0.8052,
+      "step": 2064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000193193780398018,
+      "loss": 0.7994,
+      "step": 2065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001931871828366742,
+      "loss": 0.7493,
+      "step": 2066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019318058219199371,
+      "loss": 0.7509,
+      "step": 2067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019317397846419486,
+      "loss": 0.7777,
+      "step": 2068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019316737165349618,
+      "loss": 0.7508,
+      "step": 2069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001931607617601162,
+      "loss": 0.7936,
+      "step": 2070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019315414878427373,
+      "loss": 0.782,
+      "step": 2071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019314753272618753,
+      "loss": 0.7982,
+      "step": 2072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019314091358607648,
+      "loss": 0.8014,
+      "step": 2073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019313429136415966,
+      "loss": 0.7414,
+      "step": 2074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019312766606065614,
+      "loss": 0.8067,
+      "step": 2075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019312103767578514,
+      "loss": 0.711,
+      "step": 2076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019311440620976597,
+      "loss": 0.7587,
+      "step": 2077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019310777166281807,
+      "loss": 0.817,
+      "step": 2078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001931011340351609,
+      "loss": 0.736,
+      "step": 2079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019309449332701419,
+      "loss": 0.7546,
+      "step": 2080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019308784953859758,
+      "loss": 0.7788,
+      "step": 2081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019308120267013093,
+      "loss": 0.7324,
+      "step": 2082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019307455272183415,
+      "loss": 0.7182,
+      "step": 2083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019306789969392727,
+      "loss": 0.7348,
+      "step": 2084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019306124358663045,
+      "loss": 0.8303,
+      "step": 2085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019305458440016388,
+      "loss": 0.7994,
+      "step": 2086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019304792213474797,
+      "loss": 0.7528,
+      "step": 2087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001930412567906031,
+      "loss": 0.8006,
+      "step": 2088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019303458836794982,
+      "loss": 0.7818,
+      "step": 2089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019302791686700877,
+      "loss": 0.6958,
+      "step": 2090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001930212422880007,
+      "loss": 0.7721,
+      "step": 2091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019301456463114645,
+      "loss": 0.8096,
+      "step": 2092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019300788389666696,
+      "loss": 0.7174,
+      "step": 2093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001930012000847833,
+      "loss": 0.7328,
+      "step": 2094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00019299451319571662,
+      "loss": 0.757,
+      "step": 2095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019298782322968815,
+      "loss": 0.7235,
+      "step": 2096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019298113018691927,
+      "loss": 0.784,
+      "step": 2097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019297443406763142,
+      "loss": 0.7264,
+      "step": 2098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001929677348720462,
+      "loss": 0.6451,
+      "step": 2099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019296103260038518,
+      "loss": 0.7448,
+      "step": 2100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019295432725287016,
+      "loss": 0.7604,
+      "step": 2101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019294761882972307,
+      "loss": 0.7533,
+      "step": 2102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001929409073311658,
+      "loss": 0.7541,
+      "step": 2103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019293419275742045,
+      "loss": 0.7701,
+      "step": 2104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001929274751087092,
+      "loss": 0.7758,
+      "step": 2105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019292075438525429,
+      "loss": 0.7853,
+      "step": 2106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019291403058727807,
+      "loss": 0.7657,
+      "step": 2107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001929073037150031,
+      "loss": 0.7762,
+      "step": 2108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019290057376865187,
+      "loss": 0.7164,
+      "step": 2109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019289384074844707,
+      "loss": 0.7845,
+      "step": 2110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019288710465461154,
+      "loss": 0.7078,
+      "step": 2111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001928803654873681,
+      "loss": 0.6915,
+      "step": 2112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019287362324693977,
+      "loss": 0.8249,
+      "step": 2113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001928668779335496,
+      "loss": 0.7537,
+      "step": 2114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001928601295474208,
+      "loss": 0.7784,
+      "step": 2115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019285337808877667,
+      "loss": 0.7826,
+      "step": 2116
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019284662355784055,
+      "loss": 0.7294,
+      "step": 2117
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019283986595483596,
+      "loss": 0.7721,
+      "step": 2118
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019283310527998653,
+      "loss": 0.7514,
+      "step": 2119
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001928263415335159,
+      "loss": 0.7677,
+      "step": 2120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019281957471564787,
+      "loss": 0.737,
+      "step": 2121
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019281280482660637,
+      "loss": 0.8081,
+      "step": 2122
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001928060318666154,
+      "loss": 0.7637,
+      "step": 2123
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000192799255835899,
+      "loss": 0.751,
+      "step": 2124
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019279247673468142,
+      "loss": 0.7555,
+      "step": 2125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019278569456318697,
+      "loss": 0.7022,
+      "step": 2126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019277890932164007,
+      "loss": 0.7912,
+      "step": 2127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019277212101026517,
+      "loss": 0.7762,
+      "step": 2128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001927653296292869,
+      "loss": 0.7241,
+      "step": 2129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019275853517893002,
+      "loss": 0.7781,
+      "step": 2130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019275173765941926,
+      "loss": 0.7979,
+      "step": 2131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001927449370709796,
+      "loss": 0.7815,
+      "step": 2132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019273813341383602,
+      "loss": 0.6827,
+      "step": 2133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019273132668821364,
+      "loss": 0.6898,
+      "step": 2134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019272451689433768,
+      "loss": 0.8116,
+      "step": 2135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019271770403243347,
+      "loss": 0.7178,
+      "step": 2136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019271088810272644,
+      "loss": 0.8355,
+      "step": 2137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00019270406910544208,
+      "loss": 0.6964,
+      "step": 2138
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019269724704080605,
+      "loss": 0.7233,
+      "step": 2139
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019269042190904405,
+      "loss": 0.7822,
+      "step": 2140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019268359371038188,
+      "loss": 0.769,
+      "step": 2141
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019267676244504554,
+      "loss": 0.7162,
+      "step": 2142
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019266992811326104,
+      "loss": 0.7578,
+      "step": 2143
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019266309071525446,
+      "loss": 0.7394,
+      "step": 2144
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019265625025125207,
+      "loss": 0.7886,
+      "step": 2145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019264940672148018,
+      "loss": 0.728,
+      "step": 2146
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019264256012616526,
+      "loss": 0.7666,
+      "step": 2147
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019263571046553387,
+      "loss": 0.7831,
+      "step": 2148
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019262885773981258,
+      "loss": 0.7339,
+      "step": 2149
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019262200194922815,
+      "loss": 0.6881,
+      "step": 2150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019261514309400747,
+      "loss": 0.7254,
+      "step": 2151
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001926082811743774,
+      "loss": 0.7853,
+      "step": 2152
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019260141619056507,
+      "loss": 0.7058,
+      "step": 2153
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019259454814279757,
+      "loss": 0.7257,
+      "step": 2154
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019258767703130217,
+      "loss": 0.7365,
+      "step": 2155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019258080285630618,
+      "loss": 0.7491,
+      "step": 2156
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019257392561803713,
+      "loss": 0.7358,
+      "step": 2157
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001925670453167225,
+      "loss": 0.7916,
+      "step": 2158
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019256016195258994,
+      "loss": 0.8115,
+      "step": 2159
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019255327552586727,
+      "loss": 0.7446,
+      "step": 2160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019254638603678225,
+      "loss": 0.8191,
+      "step": 2161
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019253949348556291,
+      "loss": 0.7536,
+      "step": 2162
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019253259787243733,
+      "loss": 0.8051,
+      "step": 2163
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019252569919763358,
+      "loss": 0.7769,
+      "step": 2164
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019251879746138003,
+      "loss": 0.7386,
+      "step": 2165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001925118926639049,
+      "loss": 0.7741,
+      "step": 2166
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019250498480543677,
+      "loss": 0.7853,
+      "step": 2167
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019249807388620416,
+      "loss": 0.7535,
+      "step": 2168
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019249115990643573,
+      "loss": 0.7195,
+      "step": 2169
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019248424286636028,
+      "loss": 0.76,
+      "step": 2170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019247732276620662,
+      "loss": 0.7589,
+      "step": 2171
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001924703996062038,
+      "loss": 0.7908,
+      "step": 2172
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001924634733865808,
+      "loss": 0.6949,
+      "step": 2173
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019245654410756687,
+      "loss": 0.7253,
+      "step": 2174
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019244961176939125,
+      "loss": 0.756,
+      "step": 2175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001924426763722833,
+      "loss": 0.8124,
+      "step": 2176
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019243573791647252,
+      "loss": 0.7686,
+      "step": 2177
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019242879640218844,
+      "loss": 0.7652,
+      "step": 2178
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001924218518296608,
+      "loss": 0.7634,
+      "step": 2179
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00019241490419911935,
+      "loss": 0.7455,
+      "step": 2180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000192407953510794,
+      "loss": 0.8487,
+      "step": 2181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019240099976491467,
+      "loss": 0.7851,
+      "step": 2182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019239404296171147,
+      "loss": 0.7273,
+      "step": 2183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019238708310141464,
+      "loss": 0.782,
+      "step": 2184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019238012018425436,
+      "loss": 0.7976,
+      "step": 2185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019237315421046113,
+      "loss": 0.787,
+      "step": 2186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019236618518026535,
+      "loss": 0.7123,
+      "step": 2187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001923592130938976,
+      "loss": 0.7239,
+      "step": 2188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019235223795158865,
+      "loss": 0.759,
+      "step": 2189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019234525975356923,
+      "loss": 0.7564,
+      "step": 2190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019233827850007027,
+      "loss": 0.7822,
+      "step": 2191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001923312941913227,
+      "loss": 0.7534,
+      "step": 2192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019232430682755768,
+      "loss": 0.7913,
+      "step": 2193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001923173164090064,
+      "loss": 0.7205,
+      "step": 2194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001923103229359001,
+      "loss": 0.7419,
+      "step": 2195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019230332640847024,
+      "loss": 0.7516,
+      "step": 2196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019229632682694824,
+      "loss": 0.7619,
+      "step": 2197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001922893241915658,
+      "loss": 0.7473,
+      "step": 2198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019228231850255457,
+      "loss": 0.7675,
+      "step": 2199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019227530976014632,
+      "loss": 0.7426,
+      "step": 2200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000192268297964573,
+      "loss": 0.7183,
+      "step": 2201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019226128311606658,
+      "loss": 0.7752,
+      "step": 2202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001922542652148592,
+      "loss": 0.7392,
+      "step": 2203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019224724426118305,
+      "loss": 0.7665,
+      "step": 2204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019224022025527042,
+      "loss": 0.7765,
+      "step": 2205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019223319319735375,
+      "loss": 0.8217,
+      "step": 2206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019222616308766548,
+      "loss": 0.7567,
+      "step": 2207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001922191299264383,
+      "loss": 0.7887,
+      "step": 2208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019221209371390486,
+      "loss": 0.7312,
+      "step": 2209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000192205054450298,
+      "loss": 0.7801,
+      "step": 2210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019219801213585068,
+      "loss": 0.8212,
+      "step": 2211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001921909667707958,
+      "loss": 0.756,
+      "step": 2212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001921839183553666,
+      "loss": 0.7883,
+      "step": 2213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001921768668897962,
+      "loss": 0.7603,
+      "step": 2214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019216981237431794,
+      "loss": 0.8297,
+      "step": 2215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019216275480916527,
+      "loss": 0.783,
+      "step": 2216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019215569419457165,
+      "loss": 0.7772,
+      "step": 2217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019214863053077075,
+      "loss": 0.793,
+      "step": 2218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019214156381799623,
+      "loss": 0.7583,
+      "step": 2219
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019213449405648198,
+      "loss": 0.727,
+      "step": 2220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001921274212464619,
+      "loss": 0.7294,
+      "step": 2221
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019212034538817002,
+      "loss": 0.7949,
+      "step": 2222
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019211326648184041,
+      "loss": 0.7377,
+      "step": 2223
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019210618452770732,
+      "loss": 0.7435,
+      "step": 2224
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001920990995260051,
+      "loss": 0.7814,
+      "step": 2225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001920920114769682,
+      "loss": 0.7492,
+      "step": 2226
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001920849203808311,
+      "loss": 0.7944,
+      "step": 2227
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019207782623782837,
+      "loss": 0.7605,
+      "step": 2228
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019207072904819486,
+      "loss": 0.7819,
+      "step": 2229
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019206362881216528,
+      "loss": 0.7213,
+      "step": 2230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019205652552997467,
+      "loss": 0.7116,
+      "step": 2231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000192049419201858,
+      "loss": 0.7665,
+      "step": 2232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001920423098280504,
+      "loss": 0.6887,
+      "step": 2233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019203519740878714,
+      "loss": 0.7742,
+      "step": 2234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001920280819443035,
+      "loss": 0.7006,
+      "step": 2235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019202096343483494,
+      "loss": 0.7199,
+      "step": 2236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000192013841880617,
+      "loss": 0.7312,
+      "step": 2237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001920067172818853,
+      "loss": 0.7473,
+      "step": 2238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019199958963887558,
+      "loss": 0.7768,
+      "step": 2239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001919924589518237,
+      "loss": 0.7282,
+      "step": 2240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019198532522096554,
+      "loss": 0.7692,
+      "step": 2241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019197818844653722,
+      "loss": 0.736,
+      "step": 2242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001919710486287748,
+      "loss": 0.7537,
+      "step": 2243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019196390576791458,
+      "loss": 0.7522,
+      "step": 2244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019195675986419282,
+      "loss": 0.7511,
+      "step": 2245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019194961091784606,
+      "loss": 0.7056,
+      "step": 2246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001919424589291108,
+      "loss": 0.7357,
+      "step": 2247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019193530389822363,
+      "loss": 0.6892,
+      "step": 2248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019192814582542138,
+      "loss": 0.7082,
+      "step": 2249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019192098471094085,
+      "loss": 0.7805,
+      "step": 2250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019191382055501897,
+      "loss": 0.8256,
+      "step": 2251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001919066533578928,
+      "loss": 0.8279,
+      "step": 2252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019189948311979947,
+      "loss": 0.7982,
+      "step": 2253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019189230984097627,
+      "loss": 0.7885,
+      "step": 2254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019188513352166048,
+      "loss": 0.852,
+      "step": 2255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001918779541620896,
+      "loss": 0.7729,
+      "step": 2256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001918707717625012,
+      "loss": 0.8001,
+      "step": 2257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019186358632313283,
+      "loss": 0.7827,
+      "step": 2258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019185639784422232,
+      "loss": 0.7546,
+      "step": 2259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001918492063260075,
+      "loss": 0.7255,
+      "step": 2260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019184201176872633,
+      "loss": 0.7879,
+      "step": 2261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019183481417261682,
+      "loss": 0.7443,
+      "step": 2262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019182761353791716,
+      "loss": 0.7194,
+      "step": 2263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019182040986486562,
+      "loss": 0.7487,
+      "step": 2264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019181320315370047,
+      "loss": 0.7407,
+      "step": 2265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019180599340466023,
+      "loss": 0.8548,
+      "step": 2266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019179878061798347,
+      "loss": 0.7372,
+      "step": 2267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019179156479390877,
+      "loss": 0.7863,
+      "step": 2268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019178434593267497,
+      "loss": 0.7862,
+      "step": 2269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019177712403452084,
+      "loss": 0.7879,
+      "step": 2270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019176989909968535,
+      "loss": 0.7201,
+      "step": 2271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019176267112840765,
+      "loss": 0.7382,
+      "step": 2272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019175544012092676,
+      "loss": 0.7865,
+      "step": 2273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019174820607748206,
+      "loss": 0.8055,
+      "step": 2274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001917409689983128,
+      "loss": 0.7665,
+      "step": 2275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001917337288836585,
+      "loss": 0.7348,
+      "step": 2276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019172648573375872,
+      "loss": 0.8653,
+      "step": 2277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019171923954885308,
+      "loss": 0.7339,
+      "step": 2278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019171199032918138,
+      "loss": 0.7236,
+      "step": 2279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019170473807498346,
+      "loss": 0.7209,
+      "step": 2280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019169748278649926,
+      "loss": 0.7514,
+      "step": 2281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019169022446396887,
+      "loss": 0.8477,
+      "step": 2282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019168296310763245,
+      "loss": 0.7555,
+      "step": 2283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019167569871773022,
+      "loss": 0.7448,
+      "step": 2284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019166843129450262,
+      "loss": 0.8127,
+      "step": 2285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019166116083819002,
+      "loss": 0.7805,
+      "step": 2286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019165388734903305,
+      "loss": 0.7508,
+      "step": 2287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019164661082727232,
+      "loss": 0.7373,
+      "step": 2288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019163933127314863,
+      "loss": 0.8195,
+      "step": 2289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001916320486869028,
+      "loss": 0.7166,
+      "step": 2290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019162476306877588,
+      "loss": 0.7388,
+      "step": 2291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019161747441900887,
+      "loss": 0.6967,
+      "step": 2292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019161018273784291,
+      "loss": 0.7057,
+      "step": 2293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001916028880255193,
+      "loss": 0.7019,
+      "step": 2294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001915955902822794,
+      "loss": 0.7718,
+      "step": 2295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019158828950836468,
+      "loss": 0.8137,
+      "step": 2296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001915809857040167,
+      "loss": 0.7566,
+      "step": 2297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019157367886947713,
+      "loss": 0.7996,
+      "step": 2298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019156636900498773,
+      "loss": 0.7415,
+      "step": 2299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019155905611079033,
+      "loss": 0.7282,
+      "step": 2300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019155174018712696,
+      "loss": 0.7059,
+      "step": 2301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019154442123423966,
+      "loss": 0.7441,
+      "step": 2302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019153709925237057,
+      "loss": 0.7801,
+      "step": 2303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019152977424176203,
+      "loss": 0.7954,
+      "step": 2304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001915224462026563,
+      "loss": 0.8333,
+      "step": 2305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001915151151352959,
+      "loss": 0.7804,
+      "step": 2306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001915077810399234,
+      "loss": 0.737,
+      "step": 2307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019150044391678153,
+      "loss": 0.7335,
+      "step": 2308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019149310376611296,
+      "loss": 0.7493,
+      "step": 2309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001914857605881606,
+      "loss": 0.745,
+      "step": 2310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00019147841438316737,
+      "loss": 0.7707,
+      "step": 2311
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001914710651513764,
+      "loss": 0.6553,
+      "step": 2312
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019146371289303086,
+      "loss": 0.8071,
+      "step": 2313
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019145635760837396,
+      "loss": 0.7671,
+      "step": 2314
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019144899929764912,
+      "loss": 0.7874,
+      "step": 2315
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019144163796109979,
+      "loss": 0.7484,
+      "step": 2316
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019143427359896955,
+      "loss": 0.7571,
+      "step": 2317
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019142690621150204,
+      "loss": 0.6778,
+      "step": 2318
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001914195357989411,
+      "loss": 0.7623,
+      "step": 2319
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019141216236153051,
+      "loss": 0.7815,
+      "step": 2320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019140478589951427,
+      "loss": 0.7511,
+      "step": 2321
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019139740641313643,
+      "loss": 0.7453,
+      "step": 2322
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019139002390264123,
+      "loss": 0.8154,
+      "step": 2323
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019138263836827288,
+      "loss": 0.8023,
+      "step": 2324
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019137524981027574,
+      "loss": 0.7938,
+      "step": 2325
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019136785822889435,
+      "loss": 0.7646,
+      "step": 2326
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001913604636243732,
+      "loss": 0.693,
+      "step": 2327
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019135306599695699,
+      "loss": 0.7811,
+      "step": 2328
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001913456653468905,
+      "loss": 0.7547,
+      "step": 2329
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001913382616744186,
+      "loss": 0.7319,
+      "step": 2330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019133085497978624,
+      "loss": 0.7032,
+      "step": 2331
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019132344526323852,
+      "loss": 0.6973,
+      "step": 2332
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019131603252502054,
+      "loss": 0.8191,
+      "step": 2333
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019130861676537765,
+      "loss": 0.8009,
+      "step": 2334
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019130119798455519,
+      "loss": 0.7361,
+      "step": 2335
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019129377618279863,
+      "loss": 0.7228,
+      "step": 2336
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001912863513603535,
+      "loss": 0.7693,
+      "step": 2337
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019127892351746554,
+      "loss": 0.8103,
+      "step": 2338
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019127149265438053,
+      "loss": 0.7327,
+      "step": 2339
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001912640587713442,
+      "loss": 0.7053,
+      "step": 2340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019125662186860267,
+      "loss": 0.7589,
+      "step": 2341
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019124918194640193,
+      "loss": 0.7932,
+      "step": 2342
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019124173900498818,
+      "loss": 0.7631,
+      "step": 2343
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019123429304460768,
+      "loss": 0.6995,
+      "step": 2344
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019122684406550682,
+      "loss": 0.7718,
+      "step": 2345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019121939206793203,
+      "loss": 0.7468,
+      "step": 2346
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019121193705212987,
+      "loss": 0.7886,
+      "step": 2347
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019120447901834706,
+      "loss": 0.7687,
+      "step": 2348
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019119701796683033,
+      "loss": 0.8191,
+      "step": 2349
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019118955389782656,
+      "loss": 0.7652,
+      "step": 2350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019118208681158274,
+      "loss": 0.8142,
+      "step": 2351
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001911746167083459,
+      "loss": 0.8113,
+      "step": 2352
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001911671435883632,
+      "loss": 0.8219,
+      "step": 2353
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00019115966745188198,
+      "loss": 0.7383,
+      "step": 2354
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001911521882991495,
+      "loss": 0.7295,
+      "step": 2355
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019114470613041334,
+      "loss": 0.718,
+      "step": 2356
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019113722094592098,
+      "loss": 0.7677,
+      "step": 2357
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019112973274592015,
+      "loss": 0.7173,
+      "step": 2358
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019112224153065854,
+      "loss": 0.7791,
+      "step": 2359
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019111474730038408,
+      "loss": 0.8479,
+      "step": 2360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019110725005534471,
+      "loss": 0.7677,
+      "step": 2361
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001910997497957885,
+      "loss": 0.743,
+      "step": 2362
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019109224652196362,
+      "loss": 0.7474,
+      "step": 2363
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019108474023411836,
+      "loss": 0.7801,
+      "step": 2364
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019107723093250104,
+      "loss": 0.7444,
+      "step": 2365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019106971861736013,
+      "loss": 0.8053,
+      "step": 2366
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001910622032889442,
+      "loss": 0.7483,
+      "step": 2367
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001910546849475019,
+      "loss": 0.7369,
+      "step": 2368
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019104716359328206,
+      "loss": 0.8312,
+      "step": 2369
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019103963922653344,
+      "loss": 0.6973,
+      "step": 2370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001910321118475051,
+      "loss": 0.7458,
+      "step": 2371
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019102458145644607,
+      "loss": 0.7585,
+      "step": 2372
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001910170480536055,
+      "loss": 0.7915,
+      "step": 2373
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019100951163923265,
+      "loss": 0.7015,
+      "step": 2374
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019100197221357687,
+      "loss": 0.7131,
+      "step": 2375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019099442977688766,
+      "loss": 0.7658,
+      "step": 2376
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019098688432941455,
+      "loss": 0.8613,
+      "step": 2377
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019097933587140719,
+      "loss": 0.7963,
+      "step": 2378
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001909717844031154,
+      "loss": 0.6986,
+      "step": 2379
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019096422992478896,
+      "loss": 0.7365,
+      "step": 2380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001909566724366779,
+      "loss": 0.6878,
+      "step": 2381
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019094911193903223,
+      "loss": 0.7182,
+      "step": 2382
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019094154843210211,
+      "loss": 0.7745,
+      "step": 2383
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019093398191613785,
+      "loss": 0.8337,
+      "step": 2384
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019092641239138976,
+      "loss": 0.7175,
+      "step": 2385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019091883985810832,
+      "loss": 0.7781,
+      "step": 2386
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019091126431654404,
+      "loss": 0.8033,
+      "step": 2387
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019090368576694765,
+      "loss": 0.7343,
+      "step": 2388
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019089610420956987,
+      "loss": 0.7443,
+      "step": 2389
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019088851964466154,
+      "loss": 0.7311,
+      "step": 2390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001908809320724736,
+      "loss": 0.7452,
+      "step": 2391
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019087334149325716,
+      "loss": 0.7995,
+      "step": 2392
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019086574790726335,
+      "loss": 0.7438,
+      "step": 2393
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019085815131474342,
+      "loss": 0.7235,
+      "step": 2394
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019085055171594867,
+      "loss": 0.7909,
+      "step": 2395
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019084294911113068,
+      "loss": 0.7281,
+      "step": 2396
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00019083534350054088,
+      "loss": 0.7308,
+      "step": 2397
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019082773488443097,
+      "loss": 0.7143,
+      "step": 2398
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001908201232630527,
+      "loss": 0.7765,
+      "step": 2399
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019081250863665794,
+      "loss": 0.699,
+      "step": 2400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001908048910054986,
+      "loss": 0.7443,
+      "step": 2401
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019079727036982674,
+      "loss": 0.7548,
+      "step": 2402
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001907896467298945,
+      "loss": 0.7587,
+      "step": 2403
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019078202008595416,
+      "loss": 0.7678,
+      "step": 2404
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019077439043825803,
+      "loss": 0.7737,
+      "step": 2405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019076675778705862,
+      "loss": 0.8824,
+      "step": 2406
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019075912213260837,
+      "loss": 0.7944,
+      "step": 2407
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019075148347516004,
+      "loss": 0.84,
+      "step": 2408
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019074384181496629,
+      "loss": 0.6901,
+      "step": 2409
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019073619715228,
+      "loss": 0.7929,
+      "step": 2410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019072854948735412,
+      "loss": 0.7474,
+      "step": 2411
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019072089882044168,
+      "loss": 0.7162,
+      "step": 2412
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001907132451517958,
+      "loss": 0.7121,
+      "step": 2413
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019070558848166975,
+      "loss": 0.7202,
+      "step": 2414
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019069792881031686,
+      "loss": 0.7032,
+      "step": 2415
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019069026613799058,
+      "loss": 0.7421,
+      "step": 2416
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019068260046494446,
+      "loss": 0.8043,
+      "step": 2417
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001906749317914321,
+      "loss": 0.7595,
+      "step": 2418
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019066726011770726,
+      "loss": 0.6811,
+      "step": 2419
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001906595854440238,
+      "loss": 0.7462,
+      "step": 2420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019065190777063559,
+      "loss": 0.7309,
+      "step": 2421
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019064422709779674,
+      "loss": 0.6998,
+      "step": 2422
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019063654342576136,
+      "loss": 0.739,
+      "step": 2423
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019062885675478363,
+      "loss": 0.7373,
+      "step": 2424
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000190621167085118,
+      "loss": 0.7603,
+      "step": 2425
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019061347441701876,
+      "loss": 0.7331,
+      "step": 2426
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019060577875074056,
+      "loss": 0.7231,
+      "step": 2427
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019059808008653795,
+      "loss": 0.6796,
+      "step": 2428
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019059037842466574,
+      "loss": 0.7769,
+      "step": 2429
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019058267376537868,
+      "loss": 0.8257,
+      "step": 2430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019057496610893173,
+      "loss": 0.7393,
+      "step": 2431
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019056725545557994,
+      "loss": 0.8024,
+      "step": 2432
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019055954180557843,
+      "loss": 0.7174,
+      "step": 2433
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019055182515918237,
+      "loss": 0.7713,
+      "step": 2434
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019054410551664713,
+      "loss": 0.7718,
+      "step": 2435
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019053638287822817,
+      "loss": 0.721,
+      "step": 2436
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019052865724418093,
+      "loss": 0.7431,
+      "step": 2437
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001905209286147611,
+      "loss": 0.7984,
+      "step": 2438
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00019051319699022438,
+      "loss": 0.7895,
+      "step": 2439
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001905054623708266,
+      "loss": 0.788,
+      "step": 2440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019049772475682368,
+      "loss": 0.7227,
+      "step": 2441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019048998414847158,
+      "loss": 0.802,
+      "step": 2442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001904822405460265,
+      "loss": 0.7092,
+      "step": 2443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019047449394974462,
+      "loss": 0.7984,
+      "step": 2444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019046674435988225,
+      "loss": 0.6561,
+      "step": 2445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019045899177669583,
+      "loss": 0.7057,
+      "step": 2446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019045123620044184,
+      "loss": 0.7175,
+      "step": 2447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001904434776313769,
+      "loss": 0.8014,
+      "step": 2448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019043571606975777,
+      "loss": 0.7279,
+      "step": 2449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019042795151584117,
+      "loss": 0.7454,
+      "step": 2450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019042018396988412,
+      "loss": 0.7585,
+      "step": 2451
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019041241343214352,
+      "loss": 0.6977,
+      "step": 2452
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001904046399028766,
+      "loss": 0.7108,
+      "step": 2453
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019039686338234043,
+      "loss": 0.7403,
+      "step": 2454
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019038908387079243,
+      "loss": 0.7666,
+      "step": 2455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019038130136848994,
+      "loss": 0.7684,
+      "step": 2456
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001903735158756905,
+      "loss": 0.7632,
+      "step": 2457
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001903657273926517,
+      "loss": 0.8169,
+      "step": 2458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001903579359196312,
+      "loss": 0.763,
+      "step": 2459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019035014145688689,
+      "loss": 0.8173,
+      "step": 2460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019034234400467658,
+      "loss": 0.7208,
+      "step": 2461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019033454356325833,
+      "loss": 0.7139,
+      "step": 2462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019032674013289022,
+      "loss": 0.7676,
+      "step": 2463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019031893371383043,
+      "loss": 0.8065,
+      "step": 2464
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019031112430633725,
+      "loss": 0.7545,
+      "step": 2465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019030331191066912,
+      "loss": 0.775,
+      "step": 2466
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001902954965270845,
+      "loss": 0.7265,
+      "step": 2467
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019028767815584198,
+      "loss": 0.7474,
+      "step": 2468
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019027985679720026,
+      "loss": 0.8291,
+      "step": 2469
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019027203245141814,
+      "loss": 0.7736,
+      "step": 2470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019026420511875447,
+      "loss": 0.7389,
+      "step": 2471
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001902563747994683,
+      "loss": 0.7331,
+      "step": 2472
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019024854149381865,
+      "loss": 0.728,
+      "step": 2473
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001902407052020647,
+      "loss": 0.788,
+      "step": 2474
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019023286592446584,
+      "loss": 0.718,
+      "step": 2475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019022502366128135,
+      "loss": 0.7425,
+      "step": 2476
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019021717841277073,
+      "loss": 0.7905,
+      "step": 2477
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019020933017919363,
+      "loss": 0.8064,
+      "step": 2478
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019020147896080963,
+      "loss": 0.7675,
+      "step": 2479
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019019362475787853,
+      "loss": 0.7664,
+      "step": 2480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019018576757066024,
+      "loss": 0.7771,
+      "step": 2481
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019017790739941473,
+      "loss": 0.7625,
+      "step": 2482
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019017004424440206,
+      "loss": 0.7666,
+      "step": 2483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001901621781058824,
+      "loss": 0.7864,
+      "step": 2484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019015430898411604,
+      "loss": 0.8038,
+      "step": 2485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019014643687936334,
+      "loss": 0.7475,
+      "step": 2486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019013856179188476,
+      "loss": 0.737,
+      "step": 2487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019013068372194085,
+      "loss": 0.7445,
+      "step": 2488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019012280266979235,
+      "loss": 0.7338,
+      "step": 2489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019011491863569994,
+      "loss": 0.7741,
+      "step": 2490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019010703161992455,
+      "loss": 0.7624,
+      "step": 2491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019009914162272708,
+      "loss": 0.7272,
+      "step": 2492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019009124864436864,
+      "loss": 0.6755,
+      "step": 2493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019008335268511038,
+      "loss": 0.7529,
+      "step": 2494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019007545374521355,
+      "loss": 0.7038,
+      "step": 2495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019006755182493948,
+      "loss": 0.6884,
+      "step": 2496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019005964692454965,
+      "loss": 0.7965,
+      "step": 2497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019005173904430565,
+      "loss": 0.7526,
+      "step": 2498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019004382818446907,
+      "loss": 0.7459,
+      "step": 2499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001900359143453017,
+      "loss": 0.8198,
+      "step": 2500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019002799752706537,
+      "loss": 0.7092,
+      "step": 2501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019002007773002207,
+      "loss": 0.805,
+      "step": 2502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019001215495443376,
+      "loss": 0.7484,
+      "step": 2503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019000422920056266,
+      "loss": 0.8038,
+      "step": 2504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000189996300468671,
+      "loss": 0.7678,
+      "step": 2505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001899883687590211,
+      "loss": 0.7814,
+      "step": 2506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018998043407187547,
+      "loss": 0.7282,
+      "step": 2507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018997249640749655,
+      "loss": 0.7544,
+      "step": 2508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000189964555766147,
+      "loss": 0.7147,
+      "step": 2509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018995661214808965,
+      "loss": 0.7847,
+      "step": 2510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001899486655535872,
+      "loss": 0.7867,
+      "step": 2511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018994071598290267,
+      "loss": 0.8004,
+      "step": 2512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018993276343629912,
+      "loss": 0.745,
+      "step": 2513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018992480791403958,
+      "loss": 0.713,
+      "step": 2514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018991684941638731,
+      "loss": 0.763,
+      "step": 2515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018990888794360569,
+      "loss": 0.757,
+      "step": 2516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018990092349595813,
+      "loss": 0.7844,
+      "step": 2517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001898929560737081,
+      "loss": 0.7843,
+      "step": 2518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018988498567711924,
+      "loss": 0.7436,
+      "step": 2519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018987701230645533,
+      "loss": 0.8052,
+      "step": 2520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018986903596198015,
+      "loss": 0.7334,
+      "step": 2521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018986105664395764,
+      "loss": 0.7118,
+      "step": 2522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018985307435265175,
+      "loss": 0.7917,
+      "step": 2523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018984508908832662,
+      "loss": 0.7577,
+      "step": 2524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018983710085124656,
+      "loss": 0.7676,
+      "step": 2525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018982910964167573,
+      "loss": 0.7483,
+      "step": 2526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018982111545987868,
+      "loss": 0.6609,
+      "step": 2527
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018981311830611978,
+      "loss": 0.7922,
+      "step": 2528
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018980511818066374,
+      "loss": 0.7023,
+      "step": 2529
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018979711508377523,
+      "loss": 0.7735,
+      "step": 2530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018978910901571905,
+      "loss": 0.7325,
+      "step": 2531
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001897810999767601,
+      "loss": 0.7682,
+      "step": 2532
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001897730879671634,
+      "loss": 0.7681,
+      "step": 2533
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018976507298719403,
+      "loss": 0.6692,
+      "step": 2534
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018975705503711718,
+      "loss": 0.786,
+      "step": 2535
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018974903411719815,
+      "loss": 0.7863,
+      "step": 2536
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018974101022770236,
+      "loss": 0.8318,
+      "step": 2537
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018973298336889526,
+      "loss": 0.7272,
+      "step": 2538
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018972495354104246,
+      "loss": 0.7238,
+      "step": 2539
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018971692074440963,
+      "loss": 0.8522,
+      "step": 2540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018970888497926258,
+      "loss": 0.7707,
+      "step": 2541
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001897008462458672,
+      "loss": 0.7302,
+      "step": 2542
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018969280454448945,
+      "loss": 0.6911,
+      "step": 2543
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018968475987539538,
+      "loss": 0.7006,
+      "step": 2544
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018967671223885124,
+      "loss": 0.7568,
+      "step": 2545
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018966866163512328,
+      "loss": 0.7492,
+      "step": 2546
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018966060806447784,
+      "loss": 0.6714,
+      "step": 2547
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018965255152718143,
+      "loss": 0.7477,
+      "step": 2548
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001896444920235006,
+      "loss": 0.7838,
+      "step": 2549
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018963642955370201,
+      "loss": 0.7357,
+      "step": 2550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018962836411805248,
+      "loss": 0.7361,
+      "step": 2551
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018962029571681886,
+      "loss": 0.6886,
+      "step": 2552
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018961222435026805,
+      "loss": 0.7902,
+      "step": 2553
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018960415001866718,
+      "loss": 0.7951,
+      "step": 2554
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001895960727222834,
+      "loss": 0.7456,
+      "step": 2555
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018958799246138392,
+      "loss": 0.7392,
+      "step": 2556
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018957990923623616,
+      "loss": 0.7071,
+      "step": 2557
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018957182304710755,
+      "loss": 0.7418,
+      "step": 2558
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018956373389426563,
+      "loss": 0.7706,
+      "step": 2559
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018955564177797806,
+      "loss": 0.7977,
+      "step": 2560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018954754669851258,
+      "loss": 0.7703,
+      "step": 2561
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018953944865613705,
+      "loss": 0.7982,
+      "step": 2562
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001895313476511194,
+      "loss": 0.8012,
+      "step": 2563
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018952324368372772,
+      "loss": 0.7852,
+      "step": 2564
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018951513675423006,
+      "loss": 0.7586,
+      "step": 2565
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018950702686289472,
+      "loss": 0.7283,
+      "step": 2566
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018949891400999006,
+      "loss": 0.7123,
+      "step": 2567
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018949079819578443,
+      "loss": 0.7864,
+      "step": 2568
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018948267942054647,
+      "loss": 0.766,
+      "step": 2569
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001894745576845447,
+      "loss": 0.676,
+      "step": 2570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018946643298804793,
+      "loss": 0.7828,
+      "step": 2571
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000189458305331325,
+      "loss": 0.7739,
+      "step": 2572
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001894501747146447,
+      "loss": 0.7157,
+      "step": 2573
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018944204113827622,
+      "loss": 0.7659,
+      "step": 2574
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018943390460248857,
+      "loss": 0.7387,
+      "step": 2575
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000189425765107551,
+      "loss": 0.7584,
+      "step": 2576
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018941762265373287,
+      "loss": 0.7472,
+      "step": 2577
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018940947724130353,
+      "loss": 0.8175,
+      "step": 2578
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001894013288705325,
+      "loss": 0.7549,
+      "step": 2579
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001893931775416894,
+      "loss": 0.7817,
+      "step": 2580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000189385023255044,
+      "loss": 0.7011,
+      "step": 2581
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000189376866010866,
+      "loss": 0.7521,
+      "step": 2582
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018936870580942534,
+      "loss": 0.7106,
+      "step": 2583
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018936054265099208,
+      "loss": 0.7246,
+      "step": 2584
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018935237653583623,
+      "loss": 0.7282,
+      "step": 2585
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018934420746422804,
+      "loss": 0.7372,
+      "step": 2586
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018933603543643782,
+      "loss": 0.755,
+      "step": 2587
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018932786045273593,
+      "loss": 0.8123,
+      "step": 2588
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018931968251339283,
+      "loss": 0.7333,
+      "step": 2589
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018931150161867916,
+      "loss": 0.8004,
+      "step": 2590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001893033177688656,
+      "loss": 0.7135,
+      "step": 2591
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018929513096422294,
+      "loss": 0.7942,
+      "step": 2592
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000189286941205022,
+      "loss": 0.7741,
+      "step": 2593
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018927874849153383,
+      "loss": 0.6858,
+      "step": 2594
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001892705528240295,
+      "loss": 0.743,
+      "step": 2595
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018926235420278013,
+      "loss": 0.7844,
+      "step": 2596
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018925415262805706,
+      "loss": 0.7123,
+      "step": 2597
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001892459481001316,
+      "loss": 0.7506,
+      "step": 2598
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018923774061927525,
+      "loss": 0.7026,
+      "step": 2599
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018922953018575962,
+      "loss": 0.7524,
+      "step": 2600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018922131679985626,
+      "loss": 0.8288,
+      "step": 2601
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018921310046183702,
+      "loss": 0.7124,
+      "step": 2602
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018920488117197377,
+      "loss": 0.7601,
+      "step": 2603
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018919665893053837,
+      "loss": 0.7146,
+      "step": 2604
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018918843373780296,
+      "loss": 0.741,
+      "step": 2605
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018918020559403968,
+      "loss": 0.7418,
+      "step": 2606
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018917197449952076,
+      "loss": 0.6797,
+      "step": 2607
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018916374045451853,
+      "loss": 0.7946,
+      "step": 2608
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001891555034593055,
+      "loss": 0.7425,
+      "step": 2609
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018914726351415415,
+      "loss": 0.7735,
+      "step": 2610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018913902061933714,
+      "loss": 0.7372,
+      "step": 2611
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001891307747751272,
+      "loss": 0.7474,
+      "step": 2612
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018912252598179716,
+      "loss": 0.7932,
+      "step": 2613
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018911427423961996,
+      "loss": 0.7636,
+      "step": 2614
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018910601954886866,
+      "loss": 0.7519,
+      "step": 2615
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018909776190981635,
+      "loss": 0.689,
+      "step": 2616
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018908950132273623,
+      "loss": 0.7899,
+      "step": 2617
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001890812377879017,
+      "loss": 0.8011,
+      "step": 2618
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018907297130558615,
+      "loss": 0.7624,
+      "step": 2619
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018906470187606307,
+      "loss": 0.7185,
+      "step": 2620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001890564294996061,
+      "loss": 0.7556,
+      "step": 2621
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018904815417648894,
+      "loss": 0.7336,
+      "step": 2622
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001890398759069854,
+      "loss": 0.7651,
+      "step": 2623
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001890315946913694,
+      "loss": 0.7936,
+      "step": 2624
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001890233105299149,
+      "loss": 0.7291,
+      "step": 2625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001890150234228961,
+      "loss": 0.7487,
+      "step": 2626
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018900673337058713,
+      "loss": 0.7595,
+      "step": 2627
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018899844037326225,
+      "loss": 0.7143,
+      "step": 2628
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018899014443119594,
+      "loss": 0.712,
+      "step": 2629
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018898184554466265,
+      "loss": 0.7248,
+      "step": 2630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000188973543713937,
+      "loss": 0.7641,
+      "step": 2631
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018896523893929362,
+      "loss": 0.7873,
+      "step": 2632
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018895693122100734,
+      "loss": 0.7312,
+      "step": 2633
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018894862055935302,
+      "loss": 0.7487,
+      "step": 2634
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018894030695460569,
+      "loss": 0.7508,
+      "step": 2635
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018893199040704035,
+      "loss": 0.7489,
+      "step": 2636
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001889236709169322,
+      "loss": 0.7764,
+      "step": 2637
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018891534848455654,
+      "loss": 0.7735,
+      "step": 2638
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018890702311018874,
+      "loss": 0.6876,
+      "step": 2639
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018889869479410423,
+      "loss": 0.6765,
+      "step": 2640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001888903635365786,
+      "loss": 0.7361,
+      "step": 2641
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018888202933788748,
+      "loss": 0.7048,
+      "step": 2642
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018887369219830667,
+      "loss": 0.7486,
+      "step": 2643
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018886535211811201,
+      "loss": 0.7527,
+      "step": 2644
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018885700909757944,
+      "loss": 0.8114,
+      "step": 2645
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018884866313698503,
+      "loss": 0.7701,
+      "step": 2646
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001888403142366049,
+      "loss": 0.7794,
+      "step": 2647
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018883196239671533,
+      "loss": 0.7411,
+      "step": 2648
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018882360761759262,
+      "loss": 0.7772,
+      "step": 2649
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018881524989951327,
+      "loss": 0.7166,
+      "step": 2650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018880688924275378,
+      "loss": 0.713,
+      "step": 2651
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018879852564759075,
+      "loss": 0.7208,
+      "step": 2652
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018879015911430096,
+      "loss": 0.7113,
+      "step": 2653
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018878178964316122,
+      "loss": 0.7147,
+      "step": 2654
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00018877341723444845,
+      "loss": 0.7841,
+      "step": 2655
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001887650418884397,
+      "loss": 0.7719,
+      "step": 2656
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018875666360541207,
+      "loss": 0.7652,
+      "step": 2657
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018874828238564276,
+      "loss": 0.7548,
+      "step": 2658
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018873989822940912,
+      "loss": 0.7877,
+      "step": 2659
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018873151113698854,
+      "loss": 0.7877,
+      "step": 2660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018872312110865852,
+      "loss": 0.8047,
+      "step": 2661
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018871472814469667,
+      "loss": 0.7319,
+      "step": 2662
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018870633224538072,
+      "loss": 0.7585,
+      "step": 2663
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018869793341098843,
+      "loss": 0.7639,
+      "step": 2664
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018868953164179777,
+      "loss": 0.7774,
+      "step": 2665
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018868112693808665,
+      "loss": 0.7694,
+      "step": 2666
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018867271930013315,
+      "loss": 0.7248,
+      "step": 2667
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018866430872821555,
+      "loss": 0.7212,
+      "step": 2668
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018865589522261208,
+      "loss": 0.669,
+      "step": 2669
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001886474787836011,
+      "loss": 0.796,
+      "step": 2670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018863905941146115,
+      "loss": 0.728,
+      "step": 2671
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018863063710647073,
+      "loss": 0.78,
+      "step": 2672
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001886222118689086,
+      "loss": 0.7339,
+      "step": 2673
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018861378369905348,
+      "loss": 0.7539,
+      "step": 2674
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018860535259718423,
+      "loss": 0.7712,
+      "step": 2675
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018859691856357985,
+      "loss": 0.7518,
+      "step": 2676
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018858848159851937,
+      "loss": 0.7932,
+      "step": 2677
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018858004170228197,
+      "loss": 0.763,
+      "step": 2678
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018857159887514686,
+      "loss": 0.7795,
+      "step": 2679
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018856315311739348,
+      "loss": 0.7504,
+      "step": 2680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001885547044293012,
+      "loss": 0.754,
+      "step": 2681
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018854625281114959,
+      "loss": 0.7377,
+      "step": 2682
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001885377982632183,
+      "loss": 0.7383,
+      "step": 2683
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001885293407857871,
+      "loss": 0.8184,
+      "step": 2684
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018852088037913577,
+      "loss": 0.7913,
+      "step": 2685
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018851241704354426,
+      "loss": 0.7113,
+      "step": 2686
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018850395077929258,
+      "loss": 0.7362,
+      "step": 2687
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018849548158666096,
+      "loss": 0.7868,
+      "step": 2688
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018848700946592952,
+      "loss": 0.8233,
+      "step": 2689
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001884785344173786,
+      "loss": 0.74,
+      "step": 2690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018847005644128867,
+      "loss": 0.7204,
+      "step": 2691
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001884615755379402,
+      "loss": 0.7417,
+      "step": 2692
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001884530917076138,
+      "loss": 0.7777,
+      "step": 2693
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018844460495059023,
+      "loss": 0.745,
+      "step": 2694
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001884361152671502,
+      "loss": 0.8231,
+      "step": 2695
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001884276226575747,
+      "loss": 0.7828,
+      "step": 2696
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001884191271221447,
+      "loss": 0.7596,
+      "step": 2697
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018841062866114129,
+      "loss": 0.7764,
+      "step": 2698
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001884021272748457,
+      "loss": 0.7552,
+      "step": 2699
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018839362296353915,
+      "loss": 0.8005,
+      "step": 2700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018838511572750308,
+      "loss": 0.7692,
+      "step": 2701
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018837660556701897,
+      "loss": 0.741,
+      "step": 2702
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001883680924823684,
+      "loss": 0.7961,
+      "step": 2703
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018835957647383303,
+      "loss": 0.7073,
+      "step": 2704
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018835105754169465,
+      "loss": 0.7342,
+      "step": 2705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001883425356862351,
+      "loss": 0.7689,
+      "step": 2706
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018833401090773637,
+      "loss": 0.7154,
+      "step": 2707
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018832548320648056,
+      "loss": 0.7578,
+      "step": 2708
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018831695258274977,
+      "loss": 0.7419,
+      "step": 2709
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018830841903682627,
+      "loss": 0.7751,
+      "step": 2710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018829988256899248,
+      "loss": 0.6645,
+      "step": 2711
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018829134317953074,
+      "loss": 0.6571,
+      "step": 2712
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001882828008687237,
+      "loss": 0.724,
+      "step": 2713
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018827425563685394,
+      "loss": 0.8128,
+      "step": 2714
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001882657074842042,
+      "loss": 0.7267,
+      "step": 2715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001882571564110574,
+      "loss": 0.8606,
+      "step": 2716
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018824860241769637,
+      "loss": 0.7618,
+      "step": 2717
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001882400455044042,
+      "loss": 0.7381,
+      "step": 2718
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018823148567146398,
+      "loss": 0.7735,
+      "step": 2719
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018822292291915897,
+      "loss": 0.7271,
+      "step": 2720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018821435724777249,
+      "loss": 0.7416,
+      "step": 2721
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001882057886575879,
+      "loss": 0.7122,
+      "step": 2722
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018819721714888877,
+      "loss": 0.7775,
+      "step": 2723
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018818864272195877,
+      "loss": 0.7698,
+      "step": 2724
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018818006537708147,
+      "loss": 0.7071,
+      "step": 2725
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001881714851145407,
+      "loss": 0.7937,
+      "step": 2726
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018816290193462046,
+      "loss": 0.7988,
+      "step": 2727
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001881543158376047,
+      "loss": 0.8001,
+      "step": 2728
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018814572682377746,
+      "loss": 0.8088,
+      "step": 2729
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018813713489342293,
+      "loss": 0.7383,
+      "step": 2730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018812854004682548,
+      "loss": 0.7564,
+      "step": 2731
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018811994228426943,
+      "loss": 0.7642,
+      "step": 2732
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001881113416060393,
+      "loss": 0.7347,
+      "step": 2733
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018810273801241962,
+      "loss": 0.6914,
+      "step": 2734
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001880941315036951,
+      "loss": 0.7381,
+      "step": 2735
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018808552208015046,
+      "loss": 0.6663,
+      "step": 2736
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018807690974207062,
+      "loss": 0.7588,
+      "step": 2737
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018806829448974048,
+      "loss": 0.7599,
+      "step": 2738
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018805967632344516,
+      "loss": 0.7848,
+      "step": 2739
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018805105524346977,
+      "loss": 0.7564,
+      "step": 2740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018804243125009962,
+      "loss": 0.7647,
+      "step": 2741
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018803380434362,
+      "loss": 0.7325,
+      "step": 2742
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018802517452431633,
+      "loss": 0.6924,
+      "step": 2743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018801654179247424,
+      "loss": 0.7536,
+      "step": 2744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018800790614837928,
+      "loss": 0.735,
+      "step": 2745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018799926759231723,
+      "loss": 0.7452,
+      "step": 2746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001879906261245739,
+      "loss": 0.7681,
+      "step": 2747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018798198174543523,
+      "loss": 0.7386,
+      "step": 2748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018797333445518723,
+      "loss": 0.7284,
+      "step": 2749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.000187964684254116,
+      "loss": 0.7354,
+      "step": 2750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018795603114250782,
+      "loss": 0.8015,
+      "step": 2751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001879473751206489,
+      "loss": 0.7339,
+      "step": 2752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018793871618882574,
+      "loss": 0.7591,
+      "step": 2753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018793005434732482,
+      "loss": 0.8118,
+      "step": 2754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001879213895964327,
+      "loss": 0.7239,
+      "step": 2755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018791272193643605,
+      "loss": 0.7769,
+      "step": 2756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018790405136762177,
+      "loss": 0.7714,
+      "step": 2757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018789537789027668,
+      "loss": 0.7994,
+      "step": 2758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018788670150468776,
+      "loss": 0.8188,
+      "step": 2759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001878780222111421,
+      "loss": 0.7202,
+      "step": 2760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018786934000992688,
+      "loss": 0.7731,
+      "step": 2761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001878606549013294,
+      "loss": 0.7205,
+      "step": 2762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018785196688563699,
+      "loss": 0.7614,
+      "step": 2763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001878432759631371,
+      "loss": 0.7378,
+      "step": 2764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018783458213411734,
+      "loss": 0.6818,
+      "step": 2765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018782588539886533,
+      "loss": 0.7761,
+      "step": 2766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018781718575766886,
+      "loss": 0.7645,
+      "step": 2767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018780848321081574,
+      "loss": 0.7923,
+      "step": 2768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018779977775859396,
+      "loss": 0.7385,
+      "step": 2769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018779106940129152,
+      "loss": 0.744,
+      "step": 2770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001877823581391966,
+      "loss": 0.719,
+      "step": 2771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001877736439725974,
+      "loss": 0.7687,
+      "step": 2772
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018776492690178226,
+      "loss": 0.7632,
+      "step": 2773
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001877562069270396,
+      "loss": 0.722,
+      "step": 2774
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018774748404865794,
+      "loss": 0.7738,
+      "step": 2775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018773875826692593,
+      "loss": 0.782,
+      "step": 2776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018773002958213226,
+      "loss": 0.8252,
+      "step": 2777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018772129799456572,
+      "loss": 0.7105,
+      "step": 2778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001877125635045153,
+      "loss": 0.7154,
+      "step": 2779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018770382611226987,
+      "loss": 0.7025,
+      "step": 2780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018769508581811868,
+      "loss": 0.7894,
+      "step": 2781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001876863426223508,
+      "loss": 0.6928,
+      "step": 2782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001876775965252556,
+      "loss": 0.811,
+      "step": 2783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001876688475271224,
+      "loss": 0.7152,
+      "step": 2784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018766009562824073,
+      "loss": 0.73,
+      "step": 2785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001876513408289002,
+      "loss": 0.7592,
+      "step": 2786
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001876425831293904,
+      "loss": 0.7717,
+      "step": 2787
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018763382253000118,
+      "loss": 0.7122,
+      "step": 2788
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018762505903102233,
+      "loss": 0.7252,
+      "step": 2789
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018761629263274387,
+      "loss": 0.8541,
+      "step": 2790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018760752333545587,
+      "loss": 0.7711,
+      "step": 2791
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018759875113944844,
+      "loss": 0.7931,
+      "step": 2792
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018758997604501184,
+      "loss": 0.7194,
+      "step": 2793
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018758119805243644,
+      "loss": 0.7411,
+      "step": 2794
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018757241716201267,
+      "loss": 0.7377,
+      "step": 2795
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018756363337403103,
+      "loss": 0.8007,
+      "step": 2796
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018755484668878225,
+      "loss": 0.6891,
+      "step": 2797
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018754605710655695,
+      "loss": 0.7662,
+      "step": 2798
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000187537264627646,
+      "loss": 0.7288,
+      "step": 2799
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018752846925234037,
+      "loss": 0.7282,
+      "step": 2800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000187519670980931,
+      "loss": 0.7078,
+      "step": 2801
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018751086981370907,
+      "loss": 0.68,
+      "step": 2802
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018750206575096577,
+      "loss": 0.8015,
+      "step": 2803
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018749325879299233,
+      "loss": 0.74,
+      "step": 2804
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018748444894008026,
+      "loss": 0.7707,
+      "step": 2805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000187475636192521,
+      "loss": 0.6914,
+      "step": 2806
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018746682055060612,
+      "loss": 0.8486,
+      "step": 2807
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001874580020146274,
+      "loss": 0.7958,
+      "step": 2808
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018744918058487655,
+      "loss": 0.7374,
+      "step": 2809
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018744035626164544,
+      "loss": 0.8419,
+      "step": 2810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018743152904522609,
+      "loss": 0.7086,
+      "step": 2811
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001874226989359105,
+      "loss": 0.6556,
+      "step": 2812
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018741386593399095,
+      "loss": 0.749,
+      "step": 2813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018740503003975963,
+      "loss": 0.7312,
+      "step": 2814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001873961912535089,
+      "loss": 0.7289,
+      "step": 2815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018738734957553121,
+      "loss": 0.6728,
+      "step": 2816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001873785050061191,
+      "loss": 0.7466,
+      "step": 2817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018736965754556528,
+      "loss": 0.7781,
+      "step": 2818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018736080719416242,
+      "loss": 0.7566,
+      "step": 2819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001873519539522034,
+      "loss": 0.7019,
+      "step": 2820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018734309781998114,
+      "loss": 0.763,
+      "step": 2821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018733423879778864,
+      "loss": 0.7525,
+      "step": 2822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018732537688591908,
+      "loss": 0.7294,
+      "step": 2823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018731651208466564,
+      "loss": 0.7931,
+      "step": 2824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018730764439432162,
+      "loss": 0.6998,
+      "step": 2825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018729877381518047,
+      "loss": 0.6737,
+      "step": 2826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018728990034753566,
+      "loss": 0.7545,
+      "step": 2827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018728102399168082,
+      "loss": 0.72,
+      "step": 2828
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018727214474790964,
+      "loss": 0.8348,
+      "step": 2829
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001872632626165159,
+      "loss": 0.7539,
+      "step": 2830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001872543775977935,
+      "loss": 0.7516,
+      "step": 2831
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001872454896920364,
+      "loss": 0.8144,
+      "step": 2832
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018723659889953873,
+      "loss": 0.7445,
+      "step": 2833
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018722770522059464,
+      "loss": 0.7506,
+      "step": 2834
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018721880865549837,
+      "loss": 0.7028,
+      "step": 2835
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001872099092045443,
+      "loss": 0.7723,
+      "step": 2836
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018720100686802694,
+      "loss": 0.7055,
+      "step": 2837
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018719210164624077,
+      "loss": 0.778,
+      "step": 2838
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018718319353948048,
+      "loss": 0.7677,
+      "step": 2839
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018717428254804084,
+      "loss": 0.8327,
+      "step": 2840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018716536867221666,
+      "loss": 0.7031,
+      "step": 2841
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001871564519123029,
+      "loss": 0.779,
+      "step": 2842
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018714753226859457,
+      "loss": 0.83,
+      "step": 2843
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001871386097413868,
+      "loss": 0.7042,
+      "step": 2844
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018712968433097485,
+      "loss": 0.7727,
+      "step": 2845
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000187120756037654,
+      "loss": 0.7281,
+      "step": 2846
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018711182486171968,
+      "loss": 0.8027,
+      "step": 2847
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018710289080346742,
+      "loss": 0.7092,
+      "step": 2848
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001870939538631928,
+      "loss": 0.7423,
+      "step": 2849
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018708501404119152,
+      "loss": 0.7319,
+      "step": 2850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001870760713377594,
+      "loss": 0.7341,
+      "step": 2851
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018706712575319233,
+      "loss": 0.7816,
+      "step": 2852
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018705817728778624,
+      "loss": 0.7161,
+      "step": 2853
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018704922594183732,
+      "loss": 0.7147,
+      "step": 2854
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018704027171564166,
+      "loss": 0.7794,
+      "step": 2855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018703131460949554,
+      "loss": 0.7655,
+      "step": 2856
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018702235462369538,
+      "loss": 0.7845,
+      "step": 2857
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001870133917585376,
+      "loss": 0.7648,
+      "step": 2858
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001870044260143188,
+      "loss": 0.7291,
+      "step": 2859
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018699545739133558,
+      "loss": 0.6554,
+      "step": 2860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001869864858898847,
+      "loss": 0.6833,
+      "step": 2861
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018697751151026308,
+      "loss": 0.7295,
+      "step": 2862
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018696853425276757,
+      "loss": 0.7144,
+      "step": 2863
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018695955411769525,
+      "loss": 0.72,
+      "step": 2864
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018695057110534324,
+      "loss": 0.6809,
+      "step": 2865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018694158521600875,
+      "loss": 0.7269,
+      "step": 2866
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018693259644998915,
+      "loss": 0.6987,
+      "step": 2867
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001869236048075818,
+      "loss": 0.7074,
+      "step": 2868
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018691461028908424,
+      "loss": 0.6679,
+      "step": 2869
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018690561289479404,
+      "loss": 0.7388,
+      "step": 2870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000186896612625009,
+      "loss": 0.6609,
+      "step": 2871
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001868876094800268,
+      "loss": 0.7085,
+      "step": 2872
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018687860346014537,
+      "loss": 0.7882,
+      "step": 2873
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018686959456566275,
+      "loss": 0.765,
+      "step": 2874
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018686058279687698,
+      "loss": 0.8071,
+      "step": 2875
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018685156815408625,
+      "loss": 0.705,
+      "step": 2876
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001868425506375888,
+      "loss": 0.756,
+      "step": 2877
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018683353024768302,
+      "loss": 0.7995,
+      "step": 2878
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018682450698466737,
+      "loss": 0.764,
+      "step": 2879
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018681548084884043,
+      "loss": 0.7274,
+      "step": 2880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018680645184050083,
+      "loss": 0.6974,
+      "step": 2881
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018679741995994732,
+      "loss": 0.7302,
+      "step": 2882
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018678838520747872,
+      "loss": 0.7275,
+      "step": 2883
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018677934758339401,
+      "loss": 0.7056,
+      "step": 2884
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018677030708799222,
+      "loss": 0.6716,
+      "step": 2885
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018676126372157246,
+      "loss": 0.7665,
+      "step": 2886
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018675221748443394,
+      "loss": 0.7251,
+      "step": 2887
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018674316837687598,
+      "loss": 0.7591,
+      "step": 2888
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000186734116399198,
+      "loss": 0.7454,
+      "step": 2889
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018672506155169958,
+      "loss": 0.796,
+      "step": 2890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018671600383468023,
+      "loss": 0.6744,
+      "step": 2891
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018670694324843964,
+      "loss": 0.6847,
+      "step": 2892
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001866978797932777,
+      "loss": 0.7026,
+      "step": 2893
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018668881346949417,
+      "loss": 0.7451,
+      "step": 2894
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018667974427738916,
+      "loss": 0.7908,
+      "step": 2895
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018667067221726265,
+      "loss": 0.7927,
+      "step": 2896
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018666159728941483,
+      "loss": 0.7596,
+      "step": 2897
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000186652519494146,
+      "loss": 0.7099,
+      "step": 2898
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018664343883175653,
+      "loss": 0.7592,
+      "step": 2899
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018663435530254683,
+      "loss": 0.7528,
+      "step": 2900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001866252689068175,
+      "loss": 0.7681,
+      "step": 2901
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018661617964486915,
+      "loss": 0.7014,
+      "step": 2902
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001866070875170025,
+      "loss": 0.7565,
+      "step": 2903
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001865979925235185,
+      "loss": 0.7506,
+      "step": 2904
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018658889466471794,
+      "loss": 0.6765,
+      "step": 2905
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018657979394090193,
+      "loss": 0.7498,
+      "step": 2906
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018657069035237157,
+      "loss": 0.7836,
+      "step": 2907
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018656158389942804,
+      "loss": 0.7349,
+      "step": 2908
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018655247458237276,
+      "loss": 0.8193,
+      "step": 2909
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018654336240150698,
+      "loss": 0.7461,
+      "step": 2910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001865342473571323,
+      "loss": 0.7814,
+      "step": 2911
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018652512944955033,
+      "loss": 0.7498,
+      "step": 2912
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018651600867906272,
+      "loss": 0.7133,
+      "step": 2913
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018650688504597124,
+      "loss": 0.7257,
+      "step": 2914
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018649775855057778,
+      "loss": 0.7747,
+      "step": 2915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001864886291931843,
+      "loss": 0.7656,
+      "step": 2916
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018647949697409293,
+      "loss": 0.7844,
+      "step": 2917
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018647036189360576,
+      "loss": 0.7858,
+      "step": 2918
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001864612239520251,
+      "loss": 0.729,
+      "step": 2919
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018645208314965327,
+      "loss": 0.7103,
+      "step": 2920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018644293948679274,
+      "loss": 0.7092,
+      "step": 2921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018643379296374603,
+      "loss": 0.7419,
+      "step": 2922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018642464358081579,
+      "loss": 0.8176,
+      "step": 2923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018641549133830476,
+      "loss": 0.745,
+      "step": 2924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018640633623651573,
+      "loss": 0.7331,
+      "step": 2925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018639717827575165,
+      "loss": 0.6696,
+      "step": 2926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001863880174563155,
+      "loss": 0.7619,
+      "step": 2927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018637885377851047,
+      "loss": 0.7663,
+      "step": 2928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018636968724263965,
+      "loss": 0.7723,
+      "step": 2929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018636051784900643,
+      "loss": 0.7603,
+      "step": 2930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018635134559791418,
+      "loss": 0.6911,
+      "step": 2931
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018634217048966637,
+      "loss": 0.7322,
+      "step": 2932
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018633299252456657,
+      "loss": 0.7833,
+      "step": 2933
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018632381170291851,
+      "loss": 0.8371,
+      "step": 2934
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001863146280250259,
+      "loss": 0.7388,
+      "step": 2935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018630544149119262,
+      "loss": 0.7881,
+      "step": 2936
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018629625210172268,
+      "loss": 0.7175,
+      "step": 2937
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018628705985692008,
+      "loss": 0.7704,
+      "step": 2938
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018627786475708898,
+      "loss": 0.6932,
+      "step": 2939
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018626866680253364,
+      "loss": 0.7673,
+      "step": 2940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018625946599355836,
+      "loss": 0.7585,
+      "step": 2941
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018625026233046765,
+      "loss": 0.7787,
+      "step": 2942
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018624105581356598,
+      "loss": 0.8,
+      "step": 2943
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018623184644315791,
+      "loss": 0.7003,
+      "step": 2944
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001862226342195483,
+      "loss": 0.723,
+      "step": 2945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018621341914304185,
+      "loss": 0.7217,
+      "step": 2946
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018620420121394352,
+      "loss": 0.7135,
+      "step": 2947
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018619498043255826,
+      "loss": 0.7608,
+      "step": 2948
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001861857567991912,
+      "loss": 0.7547,
+      "step": 2949
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018617653031414752,
+      "loss": 0.7755,
+      "step": 2950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001861673009777325,
+      "loss": 0.7556,
+      "step": 2951
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001861580687902515,
+      "loss": 0.7482,
+      "step": 2952
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018614883375201003,
+      "loss": 0.7831,
+      "step": 2953
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018613959586331362,
+      "loss": 0.7161,
+      "step": 2954
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018613035512446793,
+      "loss": 0.8056,
+      "step": 2955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018612111153577874,
+      "loss": 0.7846,
+      "step": 2956
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018611186509755187,
+      "loss": 0.7807,
+      "step": 2957
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018610261581009327,
+      "loss": 0.7394,
+      "step": 2958
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000186093363673709,
+      "loss": 0.7868,
+      "step": 2959
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018608410868870516,
+      "loss": 0.7461,
+      "step": 2960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000186074850855388,
+      "loss": 0.7699,
+      "step": 2961
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018606559017406382,
+      "loss": 0.7481,
+      "step": 2962
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018605632664503902,
+      "loss": 0.7853,
+      "step": 2963
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018604706026862012,
+      "loss": 0.7693,
+      "step": 2964
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018603779104511375,
+      "loss": 0.747,
+      "step": 2965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018602851897482658,
+      "loss": 0.7131,
+      "step": 2966
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018601924405806542,
+      "loss": 0.725,
+      "step": 2967
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018600996629513711,
+      "loss": 0.7377,
+      "step": 2968
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018600068568634868,
+      "loss": 0.745,
+      "step": 2969
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018599140223200716,
+      "loss": 0.8159,
+      "step": 2970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018598211593241973,
+      "loss": 0.7627,
+      "step": 2971
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018597282678789367,
+      "loss": 0.805,
+      "step": 2972
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018596353479873634,
+      "loss": 0.821,
+      "step": 2973
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018595423996525513,
+      "loss": 0.7917,
+      "step": 2974
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018594494228775768,
+      "loss": 0.7573,
+      "step": 2975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018593564176655152,
+      "loss": 0.7994,
+      "step": 2976
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018592633840194445,
+      "loss": 0.7237,
+      "step": 2977
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001859170321942443,
+      "loss": 0.749,
+      "step": 2978
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018590772314375895,
+      "loss": 0.7618,
+      "step": 2979
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018589841125079643,
+      "loss": 0.7572,
+      "step": 2980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018588909651566485,
+      "loss": 0.7698,
+      "step": 2981
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018587977893867245,
+      "loss": 0.7857,
+      "step": 2982
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018587045852012745,
+      "loss": 0.7102,
+      "step": 2983
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001858611352603383,
+      "loss": 0.7705,
+      "step": 2984
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018585180915961348,
+      "loss": 0.748,
+      "step": 2985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001858424802182615,
+      "loss": 0.7138,
+      "step": 2986
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001858331484365911,
+      "loss": 0.7793,
+      "step": 2987
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018582381381491105,
+      "loss": 0.7165,
+      "step": 2988
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001858144763535302,
+      "loss": 0.713,
+      "step": 2989
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018580513605275748,
+      "loss": 0.7344,
+      "step": 2990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018579579291290191,
+      "loss": 0.7496,
+      "step": 2991
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018578644693427273,
+      "loss": 0.736,
+      "step": 2992
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018577709811717907,
+      "loss": 0.7531,
+      "step": 2993
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018576774646193034,
+      "loss": 0.7569,
+      "step": 2994
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018575839196883592,
+      "loss": 0.7259,
+      "step": 2995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001857490346382053,
+      "loss": 0.7515,
+      "step": 2996
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001857396744703482,
+      "loss": 0.7458,
+      "step": 2997
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018573031146557422,
+      "loss": 0.6913,
+      "step": 2998
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001857209456241932,
+      "loss": 0.7728,
+      "step": 2999
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018571157694651503,
+      "loss": 0.711,
+      "step": 3000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018570220543284965,
+      "loss": 0.7363,
+      "step": 3001
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018569283108350724,
+      "loss": 0.7713,
+      "step": 3002
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018568345389879788,
+      "loss": 0.7293,
+      "step": 3003
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001856740738790319,
+      "loss": 0.7262,
+      "step": 3004
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018566469102451963,
+      "loss": 0.7865,
+      "step": 3005
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018565530533557153,
+      "loss": 0.6499,
+      "step": 3006
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018564591681249817,
+      "loss": 0.697,
+      "step": 3007
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018563652545561013,
+      "loss": 0.816,
+      "step": 3008
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018562713126521824,
+      "loss": 0.7282,
+      "step": 3009
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001856177342416333,
+      "loss": 0.6836,
+      "step": 3010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018560833438516618,
+      "loss": 0.7205,
+      "step": 3011
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018559893169612792,
+      "loss": 0.7323,
+      "step": 3012
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001855895261748297,
+      "loss": 0.7468,
+      "step": 3013
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018558011782158267,
+      "loss": 0.7366,
+      "step": 3014
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001855707066366981,
+      "loss": 0.6674,
+      "step": 3015
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018556129262048746,
+      "loss": 0.8008,
+      "step": 3016
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018555187577326215,
+      "loss": 0.7485,
+      "step": 3017
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018554245609533383,
+      "loss": 0.7843,
+      "step": 3018
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018553303358701412,
+      "loss": 0.7825,
+      "step": 3019
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018552360824861482,
+      "loss": 0.7323,
+      "step": 3020
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018551418008044777,
+      "loss": 0.7318,
+      "step": 3021
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018550474908282495,
+      "loss": 0.713,
+      "step": 3022
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018549531525605839,
+      "loss": 0.6712,
+      "step": 3023
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018548587860046023,
+      "loss": 0.7244,
+      "step": 3024
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001854764391163427,
+      "loss": 0.7297,
+      "step": 3025
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018546699680401817,
+      "loss": 0.7861,
+      "step": 3026
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000185457551663799,
+      "loss": 0.7101,
+      "step": 3027
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018544810369599776,
+      "loss": 0.7458,
+      "step": 3028
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018543865290092704,
+      "loss": 0.7588,
+      "step": 3029
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018542919927889954,
+      "loss": 0.7175,
+      "step": 3030
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018541974283022806,
+      "loss": 0.7509,
+      "step": 3031
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018541028355522547,
+      "loss": 0.6947,
+      "step": 3032
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018540082145420482,
+      "loss": 0.7074,
+      "step": 3033
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018539135652747914,
+      "loss": 0.7613,
+      "step": 3034
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018538188877536158,
+      "loss": 0.7765,
+      "step": 3035
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018537241819816544,
+      "loss": 0.6894,
+      "step": 3036
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018536294479620406,
+      "loss": 0.7317,
+      "step": 3037
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018535346856979093,
+      "loss": 0.8159,
+      "step": 3038
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018534398951923958,
+      "loss": 0.826,
+      "step": 3039
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001853345076448636,
+      "loss": 0.8552,
+      "step": 3040
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018532502294697672,
+      "loss": 0.7525,
+      "step": 3041
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018531553542589288,
+      "loss": 0.6805,
+      "step": 3042
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001853060450819259,
+      "loss": 0.8171,
+      "step": 3043
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001852965519153898,
+      "loss": 0.7255,
+      "step": 3044
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018528705592659868,
+      "loss": 0.7481,
+      "step": 3045
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018527755711586678,
+      "loss": 0.7095,
+      "step": 3046
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018526805548350837,
+      "loss": 0.7259,
+      "step": 3047
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018525855102983785,
+      "loss": 0.7159,
+      "step": 3048
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018524904375516967,
+      "loss": 0.7587,
+      "step": 3049
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018523953365981842,
+      "loss": 0.7487,
+      "step": 3050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018523002074409878,
+      "loss": 0.7455,
+      "step": 3051
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018522050500832551,
+      "loss": 0.7756,
+      "step": 3052
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018521098645281342,
+      "loss": 0.7935,
+      "step": 3053
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018520146507787746,
+      "loss": 0.6792,
+      "step": 3054
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018519194088383273,
+      "loss": 0.7672,
+      "step": 3055
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001851824138709943,
+      "loss": 0.7436,
+      "step": 3056
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001851728840396774,
+      "loss": 0.6946,
+      "step": 3057
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018516335139019742,
+      "loss": 0.7596,
+      "step": 3058
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001851538159228697,
+      "loss": 0.66,
+      "step": 3059
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018514427763800977,
+      "loss": 0.7399,
+      "step": 3060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001851347365359332,
+      "loss": 0.7089,
+      "step": 3061
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018512519261695574,
+      "loss": 0.786,
+      "step": 3062
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018511564588139312,
+      "loss": 0.7387,
+      "step": 3063
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018510609632956123,
+      "loss": 0.7514,
+      "step": 3064
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018509654396177609,
+      "loss": 0.8339,
+      "step": 3065
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018508698877835368,
+      "loss": 0.7373,
+      "step": 3066
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018507743077961024,
+      "loss": 0.7892,
+      "step": 3067
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018506786996586196,
+      "loss": 0.7497,
+      "step": 3068
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001850583063374252,
+      "loss": 0.7631,
+      "step": 3069
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018504873989461644,
+      "loss": 0.7338,
+      "step": 3070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018503917063775215,
+      "loss": 0.7509,
+      "step": 3071
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000185029598567149,
+      "loss": 0.7143,
+      "step": 3072
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018502002368312365,
+      "loss": 0.728,
+      "step": 3073
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018501044598599297,
+      "loss": 0.7956,
+      "step": 3074
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018500086547607383,
+      "loss": 0.8058,
+      "step": 3075
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018499128215368328,
+      "loss": 0.705,
+      "step": 3076
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001849816960191383,
+      "loss": 0.6978,
+      "step": 3077
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018497210707275614,
+      "loss": 0.8264,
+      "step": 3078
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018496251531485407,
+      "loss": 0.8013,
+      "step": 3079
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018495292074574947,
+      "loss": 0.7178,
+      "step": 3080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001849433233657598,
+      "loss": 0.7402,
+      "step": 3081
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018493372317520258,
+      "loss": 0.8097,
+      "step": 3082
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018492412017439552,
+      "loss": 0.8129,
+      "step": 3083
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018491451436365627,
+      "loss": 0.7787,
+      "step": 3084
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018490490574330274,
+      "loss": 0.7013,
+      "step": 3085
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018489529431365276,
+      "loss": 0.7563,
+      "step": 3086
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018488568007502448,
+      "loss": 0.747,
+      "step": 3087
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018487606302773597,
+      "loss": 0.656,
+      "step": 3088
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018486644317210535,
+      "loss": 0.791,
+      "step": 3089
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.000184856820508451,
+      "loss": 0.8001,
+      "step": 3090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018484719503709132,
+      "loss": 0.7327,
+      "step": 3091
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018483756675834477,
+      "loss": 0.6571,
+      "step": 3092
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018482793567252986,
+      "loss": 0.7404,
+      "step": 3093
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018481830177996536,
+      "loss": 0.6871,
+      "step": 3094
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018480866508097,
+      "loss": 0.7577,
+      "step": 3095
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018479902557586259,
+      "loss": 0.7103,
+      "step": 3096
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018478938326496215,
+      "loss": 0.8622,
+      "step": 3097
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018477973814858768,
+      "loss": 0.7583,
+      "step": 3098
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018477009022705825,
+      "loss": 0.7428,
+      "step": 3099
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001847604395006932,
+      "loss": 0.7857,
+      "step": 3100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018475078596981182,
+      "loss": 0.7555,
+      "step": 3101
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018474112963473352,
+      "loss": 0.6952,
+      "step": 3102
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018473147049577774,
+      "loss": 0.7424,
+      "step": 3103
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001847218085532642,
+      "loss": 0.7041,
+      "step": 3104
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018471214380751244,
+      "loss": 0.7452,
+      "step": 3105
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018470247625884234,
+      "loss": 0.7978,
+      "step": 3106
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018469280590757378,
+      "loss": 0.7358,
+      "step": 3107
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018468313275402668,
+      "loss": 0.7793,
+      "step": 3108
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018467345679852113,
+      "loss": 0.7823,
+      "step": 3109
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018466377804137726,
+      "loss": 0.7838,
+      "step": 3110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018465409648291535,
+      "loss": 0.7157,
+      "step": 3111
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018464441212345575,
+      "loss": 0.8099,
+      "step": 3112
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018463472496331885,
+      "loss": 0.7933,
+      "step": 3113
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018462503500282518,
+      "loss": 0.714,
+      "step": 3114
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018461534224229538,
+      "loss": 0.7511,
+      "step": 3115
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018460564668205013,
+      "loss": 0.7186,
+      "step": 3116
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018459594832241028,
+      "loss": 0.8094,
+      "step": 3117
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018458624716369667,
+      "loss": 0.7805,
+      "step": 3118
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001845765432062303,
+      "loss": 0.706,
+      "step": 3119
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001845668364503323,
+      "loss": 0.7789,
+      "step": 3120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018455712689632376,
+      "loss": 0.6749,
+      "step": 3121
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018454741454452603,
+      "loss": 0.7869,
+      "step": 3122
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018453769939526043,
+      "loss": 0.8077,
+      "step": 3123
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018452798144884838,
+      "loss": 0.7123,
+      "step": 3124
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018451826070561147,
+      "loss": 0.7277,
+      "step": 3125
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018450853716587133,
+      "loss": 0.7289,
+      "step": 3126
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018449881082994969,
+      "loss": 0.7799,
+      "step": 3127
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018448908169816833,
+      "loss": 0.7729,
+      "step": 3128
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001844793497708492,
+      "loss": 0.8341,
+      "step": 3129
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001844696150483143,
+      "loss": 0.7014,
+      "step": 3130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001844598775308857,
+      "loss": 0.7581,
+      "step": 3131
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018445013721888565,
+      "loss": 0.7626,
+      "step": 3132
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018444039411263637,
+      "loss": 0.7183,
+      "step": 3133
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018443064821246026,
+      "loss": 0.7394,
+      "step": 3134
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001844208995186798,
+      "loss": 0.7186,
+      "step": 3135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018441114803161756,
+      "loss": 0.7787,
+      "step": 3136
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001844013937515962,
+      "loss": 0.6928,
+      "step": 3137
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001843916366789384,
+      "loss": 0.7458,
+      "step": 3138
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018438187681396704,
+      "loss": 0.6755,
+      "step": 3139
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018437211415700509,
+      "loss": 0.7928,
+      "step": 3140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018436234870837547,
+      "loss": 0.7092,
+      "step": 3141
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001843525804684014,
+      "loss": 0.739,
+      "step": 3142
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018434280943740606,
+      "loss": 0.7479,
+      "step": 3143
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001843330356157127,
+      "loss": 0.7729,
+      "step": 3144
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018432325900364472,
+      "loss": 0.7235,
+      "step": 3145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001843134796015257,
+      "loss": 0.7518,
+      "step": 3146
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001843036974096791,
+      "loss": 0.7102,
+      "step": 3147
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018429391242842863,
+      "loss": 0.724,
+      "step": 3148
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018428412465809808,
+      "loss": 0.7755,
+      "step": 3149
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018427433409901127,
+      "loss": 0.8609,
+      "step": 3150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018426454075149215,
+      "loss": 0.7312,
+      "step": 3151
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018425474461586475,
+      "loss": 0.8106,
+      "step": 3152
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018424494569245327,
+      "loss": 0.7725,
+      "step": 3153
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018423514398158183,
+      "loss": 0.7419,
+      "step": 3154
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018422533948357482,
+      "loss": 0.7596,
+      "step": 3155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018421553219875658,
+      "loss": 0.7291,
+      "step": 3156
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018420572212745167,
+      "loss": 0.7618,
+      "step": 3157
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018419590926998467,
+      "loss": 0.7496,
+      "step": 3158
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018418609362668025,
+      "loss": 0.6783,
+      "step": 3159
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018417627519786315,
+      "loss": 0.8029,
+      "step": 3160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001841664539838583,
+      "loss": 0.6788,
+      "step": 3161
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018415662998499065,
+      "loss": 0.7182,
+      "step": 3162
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001841468032015852,
+      "loss": 0.7724,
+      "step": 3163
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018413697363396718,
+      "loss": 0.7129,
+      "step": 3164
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018412714128246174,
+      "loss": 0.7764,
+      "step": 3165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018411730614739427,
+      "loss": 0.7699,
+      "step": 3166
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018410746822909016,
+      "loss": 0.8287,
+      "step": 3167
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018409762752787497,
+      "loss": 0.717,
+      "step": 3168
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001840877840440742,
+      "loss": 0.7153,
+      "step": 3169
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018407793777801366,
+      "loss": 0.7949,
+      "step": 3170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018406808873001906,
+      "loss": 0.7703,
+      "step": 3171
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018405823690041632,
+      "loss": 0.7039,
+      "step": 3172
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001840483822895314,
+      "loss": 0.6974,
+      "step": 3173
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001840385248976904,
+      "loss": 0.7298,
+      "step": 3174
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018402866472521946,
+      "loss": 0.7287,
+      "step": 3175
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018401880177244478,
+      "loss": 0.7334,
+      "step": 3176
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018400893603969274,
+      "loss": 0.8105,
+      "step": 3177
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018399906752728976,
+      "loss": 0.7372,
+      "step": 3178
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018398919623556238,
+      "loss": 0.6793,
+      "step": 3179
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018397932216483723,
+      "loss": 0.7411,
+      "step": 3180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000183969445315441,
+      "loss": 0.8084,
+      "step": 3181
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018395956568770046,
+      "loss": 0.7849,
+      "step": 3182
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018394968328194257,
+      "loss": 0.7622,
+      "step": 3183
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018393979809849426,
+      "loss": 0.6798,
+      "step": 3184
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018392991013768266,
+      "loss": 0.767,
+      "step": 3185
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018392001939983486,
+      "loss": 0.7503,
+      "step": 3186
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018391012588527821,
+      "loss": 0.7749,
+      "step": 3187
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018390022959433998,
+      "loss": 0.6908,
+      "step": 3188
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018389033052734768,
+      "loss": 0.7082,
+      "step": 3189
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001838804286846288,
+      "loss": 0.7665,
+      "step": 3190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018387052406651099,
+      "loss": 0.7355,
+      "step": 3191
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018386061667332196,
+      "loss": 0.7299,
+      "step": 3192
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018385070650538956,
+      "loss": 0.7436,
+      "step": 3193
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018384079356304162,
+      "loss": 0.7127,
+      "step": 3194
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001838308778466062,
+      "loss": 0.8077,
+      "step": 3195
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018382095935641135,
+      "loss": 0.758,
+      "step": 3196
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018381103809278525,
+      "loss": 0.7374,
+      "step": 3197
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001838011140560562,
+      "loss": 0.7561,
+      "step": 3198
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018379118724655254,
+      "loss": 0.6834,
+      "step": 3199
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018378125766460273,
+      "loss": 0.8167,
+      "step": 3200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018377132531053532,
+      "loss": 0.7173,
+      "step": 3201
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018376139018467892,
+      "loss": 0.703,
+      "step": 3202
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018375145228736228,
+      "loss": 0.7731,
+      "step": 3203
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018374151161891424,
+      "loss": 0.7822,
+      "step": 3204
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018373156817966368,
+      "loss": 0.7557,
+      "step": 3205
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018372162196993963,
+      "loss": 0.7694,
+      "step": 3206
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018371167299007115,
+      "loss": 0.7367,
+      "step": 3207
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018370172124038746,
+      "loss": 0.7734,
+      "step": 3208
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018369176672121783,
+      "loss": 0.7621,
+      "step": 3209
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018368180943289164,
+      "loss": 0.714,
+      "step": 3210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018367184937573833,
+      "loss": 0.773,
+      "step": 3211
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018366188655008747,
+      "loss": 0.7785,
+      "step": 3212
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018365192095626872,
+      "loss": 0.7741,
+      "step": 3213
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001836419525946118,
+      "loss": 0.7656,
+      "step": 3214
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001836319814654465,
+      "loss": 0.7722,
+      "step": 3215
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018362200756910283,
+      "loss": 0.6888,
+      "step": 3216
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018361203090591071,
+      "loss": 0.7406,
+      "step": 3217
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018360205147620028,
+      "loss": 0.766,
+      "step": 3218
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018359206928030174,
+      "loss": 0.7452,
+      "step": 3219
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001835820843185454,
+      "loss": 0.7862,
+      "step": 3220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001835720965912616,
+      "loss": 0.7738,
+      "step": 3221
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018356210609878085,
+      "loss": 0.7514,
+      "step": 3222
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018355211284143365,
+      "loss": 0.7919,
+      "step": 3223
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018354211681955072,
+      "loss": 0.7918,
+      "step": 3224
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018353211803346273,
+      "loss": 0.7169,
+      "step": 3225
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001835221164835006,
+      "loss": 0.7424,
+      "step": 3226
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018351211216999518,
+      "loss": 0.752,
+      "step": 3227
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018350210509327752,
+      "loss": 0.7189,
+      "step": 3228
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018349209525367878,
+      "loss": 0.6811,
+      "step": 3229
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018348208265153005,
+      "loss": 0.7502,
+      "step": 3230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018347206728716272,
+      "loss": 0.668,
+      "step": 3231
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018346204916090813,
+      "loss": 0.7116,
+      "step": 3232
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018345202827309775,
+      "loss": 0.7395,
+      "step": 3233
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018344200462406315,
+      "loss": 0.7356,
+      "step": 3234
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018343197821413604,
+      "loss": 0.7623,
+      "step": 3235
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018342194904364813,
+      "loss": 0.7524,
+      "step": 3236
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018341191711293125,
+      "loss": 0.7525,
+      "step": 3237
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018340188242231732,
+      "loss": 0.86,
+      "step": 3238
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018339184497213838,
+      "loss": 0.7463,
+      "step": 3239
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018338180476272658,
+      "loss": 0.7099,
+      "step": 3240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018337176179441408,
+      "loss": 0.6992,
+      "step": 3241
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001833617160675332,
+      "loss": 0.7267,
+      "step": 3242
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018335166758241632,
+      "loss": 0.7848,
+      "step": 3243
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018334161633939595,
+      "loss": 0.8136,
+      "step": 3244
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001833315623388046,
+      "loss": 0.7569,
+      "step": 3245
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000183321505580975,
+      "loss": 0.7589,
+      "step": 3246
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018331144606623982,
+      "loss": 0.6713,
+      "step": 3247
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018330138379493195,
+      "loss": 0.6906,
+      "step": 3248
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018329131876738437,
+      "loss": 0.7703,
+      "step": 3249
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018328125098393005,
+      "loss": 0.6756,
+      "step": 3250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018327118044490212,
+      "loss": 0.7421,
+      "step": 3251
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018326110715063382,
+      "loss": 0.7301,
+      "step": 3252
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001832510311014584,
+      "loss": 0.7537,
+      "step": 3253
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001832409522977093,
+      "loss": 0.7147,
+      "step": 3254
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018323087073971993,
+      "loss": 0.7571,
+      "step": 3255
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018322078642782396,
+      "loss": 0.749,
+      "step": 3256
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018321069936235503,
+      "loss": 0.7395,
+      "step": 3257
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018320060954364685,
+      "loss": 0.7378,
+      "step": 3258
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018319051697203328,
+      "loss": 0.724,
+      "step": 3259
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018318042164784827,
+      "loss": 0.7515,
+      "step": 3260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001831703235714259,
+      "loss": 0.7396,
+      "step": 3261
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001831602227431002,
+      "loss": 0.7292,
+      "step": 3262
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018315011916320543,
+      "loss": 0.7259,
+      "step": 3263
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018314001283207588,
+      "loss": 0.753,
+      "step": 3264
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018312990375004597,
+      "loss": 0.7303,
+      "step": 3265
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018311979191745016,
+      "loss": 0.7532,
+      "step": 3266
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018310967733462303,
+      "loss": 0.7315,
+      "step": 3267
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018309956000189928,
+      "loss": 0.7992,
+      "step": 3268
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001830894399196136,
+      "loss": 0.7689,
+      "step": 3269
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001830793170881009,
+      "loss": 0.7716,
+      "step": 3270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018306919150769607,
+      "loss": 0.7065,
+      "step": 3271
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018305906317873417,
+      "loss": 0.7947,
+      "step": 3272
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018304893210155035,
+      "loss": 0.7081,
+      "step": 3273
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018303879827647975,
+      "loss": 0.7546,
+      "step": 3274
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018302866170385776,
+      "loss": 0.8019,
+      "step": 3275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018301852238401968,
+      "loss": 0.7319,
+      "step": 3276
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018300838031730108,
+      "loss": 0.727,
+      "step": 3277
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001829982355040375,
+      "loss": 0.7272,
+      "step": 3278
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018298808794456458,
+      "loss": 0.7706,
+      "step": 3279
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018297793763921813,
+      "loss": 0.7599,
+      "step": 3280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018296778458833397,
+      "loss": 0.8141,
+      "step": 3281
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018295762879224804,
+      "loss": 0.7144,
+      "step": 3282
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018294747025129638,
+      "loss": 0.7072,
+      "step": 3283
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018293730896581508,
+      "loss": 0.7029,
+      "step": 3284
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001829271449361404,
+      "loss": 0.7552,
+      "step": 3285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018291697816260865,
+      "loss": 0.7599,
+      "step": 3286
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018290680864555616,
+      "loss": 0.717,
+      "step": 3287
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018328960423062528,
+      "loss": 0.6724,
+      "step": 3288
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018327965580617342,
+      "loss": 0.7441,
+      "step": 3289
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018326970469139192,
+      "loss": 0.73,
+      "step": 3290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018325975088660218,
+      "loss": 0.6981,
+      "step": 3291
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018324979439212583,
+      "loss": 0.7608,
+      "step": 3292
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018323983520828442,
+      "loss": 0.6645,
+      "step": 3293
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018322987333539977,
+      "loss": 0.7299,
+      "step": 3294
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018321990877379372,
+      "loss": 0.7406,
+      "step": 3295
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018320994152378807,
+      "loss": 0.7458,
+      "step": 3296
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001831999715857048,
+      "loss": 0.7276,
+      "step": 3297
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018318999895986613,
+      "loss": 0.7918,
+      "step": 3298
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001831800236465941,
+      "loss": 0.678,
+      "step": 3299
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018317004564621096,
+      "loss": 0.6754,
+      "step": 3300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018316006495903913,
+      "loss": 0.7356,
+      "step": 3301
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018315008158540096,
+      "loss": 0.7184,
+      "step": 3302
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018314009552561898,
+      "loss": 0.7164,
+      "step": 3303
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001831301067800158,
+      "loss": 0.7112,
+      "step": 3304
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018312011534891412,
+      "loss": 0.758,
+      "step": 3305
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018311012123263667,
+      "loss": 0.7311,
+      "step": 3306
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001831001244315063,
+      "loss": 0.6739,
+      "step": 3307
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000183090124945846,
+      "loss": 0.8239,
+      "step": 3308
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018308012277597883,
+      "loss": 0.7711,
+      "step": 3309
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001830701179222278,
+      "loss": 0.7761,
+      "step": 3310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018306011038491619,
+      "loss": 0.7315,
+      "step": 3311
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018305010016436726,
+      "loss": 0.715,
+      "step": 3312
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018304008726090443,
+      "loss": 0.7413,
+      "step": 3313
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001830300716748511,
+      "loss": 0.7427,
+      "step": 3314
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018302005340653087,
+      "loss": 0.6848,
+      "step": 3315
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018301003245626738,
+      "loss": 0.7067,
+      "step": 3316
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018300000882438434,
+      "loss": 0.7753,
+      "step": 3317
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018298998251120552,
+      "loss": 0.7454,
+      "step": 3318
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018297995351705492,
+      "loss": 0.8116,
+      "step": 3319
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018296992184225642,
+      "loss": 0.7204,
+      "step": 3320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018295988748713416,
+      "loss": 0.7121,
+      "step": 3321
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018294985045201227,
+      "loss": 0.7969,
+      "step": 3322
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018293981073721498,
+      "loss": 0.7602,
+      "step": 3323
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018292976834306663,
+      "loss": 0.763,
+      "step": 3324
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018291972326989165,
+      "loss": 0.7819,
+      "step": 3325
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018290967551801455,
+      "loss": 0.6559,
+      "step": 3326
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001828996250877599,
+      "loss": 0.7158,
+      "step": 3327
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018288957197945237,
+      "loss": 0.7867,
+      "step": 3328
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018287951619341678,
+      "loss": 0.7451,
+      "step": 3329
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001828694577299779,
+      "loss": 0.7483,
+      "step": 3330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001828593965894607,
+      "loss": 0.7247,
+      "step": 3331
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001828493327721902,
+      "loss": 0.769,
+      "step": 3332
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018283926627849154,
+      "loss": 0.8108,
+      "step": 3333
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001828291971086899,
+      "loss": 0.7641,
+      "step": 3334
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018281912526311054,
+      "loss": 0.7479,
+      "step": 3335
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018280905074207884,
+      "loss": 0.8065,
+      "step": 3336
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018279897354592025,
+      "loss": 0.7116,
+      "step": 3337
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001827888936749603,
+      "loss": 0.7287,
+      "step": 3338
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001827788111295247,
+      "loss": 0.7357,
+      "step": 3339
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018276872590993902,
+      "loss": 0.7016,
+      "step": 3340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001827586380165292,
+      "loss": 0.7086,
+      "step": 3341
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018274854744962103,
+      "loss": 0.8487,
+      "step": 3342
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018273845420954056,
+      "loss": 0.6508,
+      "step": 3343
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018272835829661376,
+      "loss": 0.7377,
+      "step": 3344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018271825971116686,
+      "loss": 0.8693,
+      "step": 3345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018270815845352608,
+      "loss": 0.7397,
+      "step": 3346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001826980545240177,
+      "loss": 0.7549,
+      "step": 3347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018268794792296813,
+      "loss": 0.7354,
+      "step": 3348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018267783865070388,
+      "loss": 0.7251,
+      "step": 3349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018266772670755152,
+      "loss": 0.7397,
+      "step": 3350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018265761209383774,
+      "loss": 0.7254,
+      "step": 3351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018264749480988924,
+      "loss": 0.678,
+      "step": 3352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018263737485603289,
+      "loss": 0.7455,
+      "step": 3353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018262725223259563,
+      "loss": 0.6759,
+      "step": 3354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001826171269399044,
+      "loss": 0.7407,
+      "step": 3355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018260699897828636,
+      "loss": 0.7502,
+      "step": 3356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018259686834806867,
+      "loss": 0.7004,
+      "step": 3357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018258673504957865,
+      "loss": 0.7524,
+      "step": 3358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018257659908314352,
+      "loss": 0.8022,
+      "step": 3359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018256646044909086,
+      "loss": 0.7713,
+      "step": 3360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018255631914774808,
+      "loss": 0.7769,
+      "step": 3361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018254617517944293,
+      "loss": 0.7183,
+      "step": 3362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018253602854450299,
+      "loss": 0.7442,
+      "step": 3363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001825258792432561,
+      "loss": 0.7573,
+      "step": 3364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018251572727603005,
+      "loss": 0.8268,
+      "step": 3365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018250557264315294,
+      "loss": 0.7255,
+      "step": 3366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018249541534495268,
+      "loss": 0.757,
+      "step": 3367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001824852553817575,
+      "loss": 0.6909,
+      "step": 3368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018247509275389555,
+      "loss": 0.7323,
+      "step": 3369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018246492746169514,
+      "loss": 0.7232,
+      "step": 3370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001824547595054847,
+      "loss": 0.7326,
+      "step": 3371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018244458888559266,
+      "loss": 0.83,
+      "step": 3372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018243441560234758,
+      "loss": 0.6386,
+      "step": 3373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018242423965607812,
+      "loss": 0.7591,
+      "step": 3374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000182414061047113,
+      "loss": 0.7766,
+      "step": 3375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018240387977578107,
+      "loss": 0.7966,
+      "step": 3376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018239369584241118,
+      "loss": 0.7229,
+      "step": 3377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018238350924733236,
+      "loss": 0.7072,
+      "step": 3378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001823733199908737,
+      "loss": 0.6916,
+      "step": 3379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018236312807336428,
+      "loss": 0.8907,
+      "step": 3380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018235293349513345,
+      "loss": 0.75,
+      "step": 3381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018234273625651048,
+      "loss": 0.7275,
+      "step": 3382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001823325363578248,
+      "loss": 0.7414,
+      "step": 3383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018232233379940596,
+      "loss": 0.7488,
+      "step": 3384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001823121285815835,
+      "loss": 0.7163,
+      "step": 3385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018230192070468708,
+      "loss": 0.7386,
+      "step": 3386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018229171016904651,
+      "loss": 0.6998,
+      "step": 3387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001822814969749916,
+      "loss": 0.7712,
+      "step": 3388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018227128112285234,
+      "loss": 0.7904,
+      "step": 3389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001822610626129587,
+      "loss": 0.7423,
+      "step": 3390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018225084144564085,
+      "loss": 0.7556,
+      "step": 3391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001822406176212289,
+      "loss": 0.7347,
+      "step": 3392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018223039114005315,
+      "loss": 0.6986,
+      "step": 3393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018222016200244396,
+      "loss": 0.78,
+      "step": 3394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018220993020873184,
+      "loss": 0.7407,
+      "step": 3395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001821996957592473,
+      "loss": 0.7271,
+      "step": 3396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001821894586543209,
+      "loss": 0.6759,
+      "step": 3397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018217921889428343,
+      "loss": 0.7515,
+      "step": 3398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001821689764794656,
+      "loss": 0.6696,
+      "step": 3399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018215873141019836,
+      "loss": 0.748,
+      "step": 3400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018214848368681267,
+      "loss": 0.7168,
+      "step": 3401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018213823330963955,
+      "loss": 0.7278,
+      "step": 3402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018212798027901012,
+      "loss": 0.7269,
+      "step": 3403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018211772459525565,
+      "loss": 0.7739,
+      "step": 3404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018210746625870746,
+      "loss": 0.7449,
+      "step": 3405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018209720526969684,
+      "loss": 0.6946,
+      "step": 3406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018208694162855538,
+      "loss": 0.8191,
+      "step": 3407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001820766753356146,
+      "loss": 0.7374,
+      "step": 3408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018206640639120613,
+      "loss": 0.7872,
+      "step": 3409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018205613479566174,
+      "loss": 0.7604,
+      "step": 3410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018204586054931324,
+      "loss": 0.6984,
+      "step": 3411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018203558365249253,
+      "loss": 0.7657,
+      "step": 3412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018202530410553163,
+      "loss": 0.7494,
+      "step": 3413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001820150219087626,
+      "loss": 0.7854,
+      "step": 3414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018200473706251756,
+      "loss": 0.6754,
+      "step": 3415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018199444956712882,
+      "loss": 0.734,
+      "step": 3416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001819841594229287,
+      "loss": 0.7456,
+      "step": 3417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018197386663024962,
+      "loss": 0.7514,
+      "step": 3418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001819635711894241,
+      "loss": 0.7562,
+      "step": 3419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018195327310078466,
+      "loss": 0.7615,
+      "step": 3420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001819429723646641,
+      "loss": 0.6933,
+      "step": 3421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018193266898139507,
+      "loss": 0.7141,
+      "step": 3422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018192236295131047,
+      "loss": 0.746,
+      "step": 3423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018191205427474323,
+      "loss": 0.7665,
+      "step": 3424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018190174295202636,
+      "loss": 0.8173,
+      "step": 3425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018189142898349299,
+      "loss": 0.7141,
+      "step": 3426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018188111236947628,
+      "loss": 0.8058,
+      "step": 3427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001818707931103095,
+      "loss": 0.7152,
+      "step": 3428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018186047120632604,
+      "loss": 0.727,
+      "step": 3429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018185014665785936,
+      "loss": 0.7353,
+      "step": 3430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018183981946524297,
+      "loss": 0.7716,
+      "step": 3431
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018182948962881045,
+      "loss": 0.7405,
+      "step": 3432
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018181915714889554,
+      "loss": 0.689,
+      "step": 3433
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018180882202583205,
+      "loss": 0.7522,
+      "step": 3434
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018179848425995383,
+      "loss": 0.7516,
+      "step": 3435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018178814385159478,
+      "loss": 0.689,
+      "step": 3436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001817778008010891,
+      "loss": 0.707,
+      "step": 3437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018176745510877073,
+      "loss": 0.7012,
+      "step": 3438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018175710677497404,
+      "loss": 0.7356,
+      "step": 3439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018174675580003325,
+      "loss": 0.7405,
+      "step": 3440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018173640218428275,
+      "loss": 0.7069,
+      "step": 3441
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018172604592805707,
+      "loss": 0.7742,
+      "step": 3442
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001817156870316907,
+      "loss": 0.7203,
+      "step": 3443
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018170532549551828,
+      "loss": 0.7576,
+      "step": 3444
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018169496131987457,
+      "loss": 0.777,
+      "step": 3445
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001816845945050944,
+      "loss": 0.7746,
+      "step": 3446
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001816742250515126,
+      "loss": 0.7732,
+      "step": 3447
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018166385295946422,
+      "loss": 0.7033,
+      "step": 3448
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018165347822928432,
+      "loss": 0.7828,
+      "step": 3449
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.000181643100861308,
+      "loss": 0.7856,
+      "step": 3450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018163272085587056,
+      "loss": 0.7975,
+      "step": 3451
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018162233821330729,
+      "loss": 0.7597,
+      "step": 3452
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001816119529339536,
+      "loss": 0.7631,
+      "step": 3453
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.000181601565018145,
+      "loss": 0.7496,
+      "step": 3454
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018159117446621702,
+      "loss": 0.6918,
+      "step": 3455
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018158078127850538,
+      "loss": 0.7751,
+      "step": 3456
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018157038545534585,
+      "loss": 0.7606,
+      "step": 3457
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018155998699707416,
+      "loss": 0.6861,
+      "step": 3458
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018154958590402637,
+      "loss": 0.7342,
+      "step": 3459
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018153918217653835,
+      "loss": 0.8037,
+      "step": 3460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001815287758149463,
+      "loss": 0.8329,
+      "step": 3461
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018151836681958632,
+      "loss": 0.7498,
+      "step": 3462
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018150795519079467,
+      "loss": 0.7447,
+      "step": 3463
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001814975409289078,
+      "loss": 0.7724,
+      "step": 3464
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.000181487124034262,
+      "loss": 0.7118,
+      "step": 3465
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018147670450719384,
+      "loss": 0.6736,
+      "step": 3466
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018146628234803997,
+      "loss": 0.762,
+      "step": 3467
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018145585755713703,
+      "loss": 0.7416,
+      "step": 3468
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018144543013482177,
+      "loss": 0.6831,
+      "step": 3469
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018143500008143112,
+      "loss": 0.7969,
+      "step": 3470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001814245673973019,
+      "loss": 0.7119,
+      "step": 3471
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018141413208277123,
+      "loss": 0.6822,
+      "step": 3472
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018140369413817623,
+      "loss": 0.754,
+      "step": 3473
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.000181393253563854,
+      "loss": 0.7089,
+      "step": 3474
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001813828103601419,
+      "loss": 0.7506,
+      "step": 3475
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018137236452737732,
+      "loss": 0.7381,
+      "step": 3476
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018136191606589762,
+      "loss": 0.6867,
+      "step": 3477
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001813514649760404,
+      "loss": 0.7523,
+      "step": 3478
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018134101125814327,
+      "loss": 0.7483,
+      "step": 3479
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018133055491254392,
+      "loss": 0.7221,
+      "step": 3480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018132009593958015,
+      "loss": 0.7273,
+      "step": 3481
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018130963433958983,
+      "loss": 0.695,
+      "step": 3482
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001812991701129109,
+      "loss": 0.7667,
+      "step": 3483
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018128870325988144,
+      "loss": 0.7763,
+      "step": 3484
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018127823378083954,
+      "loss": 0.7098,
+      "step": 3485
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001812677616761235,
+      "loss": 0.7441,
+      "step": 3486
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001812572869460715,
+      "loss": 0.6937,
+      "step": 3487
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.000181246809591022,
+      "loss": 0.7418,
+      "step": 3488
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018123632961131346,
+      "loss": 0.6869,
+      "step": 3489
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018122584700728443,
+      "loss": 0.7095,
+      "step": 3490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018121536177927348,
+      "loss": 0.7606,
+      "step": 3491
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018120487392761945,
+      "loss": 0.7558,
+      "step": 3492
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001811943834526611,
+      "loss": 0.7459,
+      "step": 3493
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018118389035473727,
+      "loss": 0.7361,
+      "step": 3494
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.000181173394634187,
+      "loss": 0.758,
+      "step": 3495
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018116289629134934,
+      "loss": 0.7381,
+      "step": 3496
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018115239532656344,
+      "loss": 0.7154,
+      "step": 3497
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018114189174016848,
+      "loss": 0.6799,
+      "step": 3498
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018113138553250383,
+      "loss": 0.7157,
+      "step": 3499
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001811208767039089,
+      "loss": 0.7148,
+      "step": 3500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018111036525472312,
+      "loss": 0.7223,
+      "step": 3501
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018109985118528612,
+      "loss": 0.777,
+      "step": 3502
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001810893344959375,
+      "loss": 0.7228,
+      "step": 3503
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018107881518701703,
+      "loss": 0.7237,
+      "step": 3504
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018106829325886453,
+      "loss": 0.7747,
+      "step": 3505
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018105776871181993,
+      "loss": 0.74,
+      "step": 3506
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018104724154622317,
+      "loss": 0.6622,
+      "step": 3507
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018103671176241437,
+      "loss": 0.7018,
+      "step": 3508
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001810261793607337,
+      "loss": 0.6833,
+      "step": 3509
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018101564434152136,
+      "loss": 0.7112,
+      "step": 3510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018100510670511773,
+      "loss": 0.7098,
+      "step": 3511
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001809945664518632,
+      "loss": 0.7525,
+      "step": 3512
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018098402358209824,
+      "loss": 0.7854,
+      "step": 3513
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001809734780961635,
+      "loss": 0.7937,
+      "step": 3514
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001809629299943996,
+      "loss": 0.775,
+      "step": 3515
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018095237927714735,
+      "loss": 0.7766,
+      "step": 3516
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018094182594474752,
+      "loss": 0.7746,
+      "step": 3517
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001809312699975411,
+      "loss": 0.7214,
+      "step": 3518
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.000180920711435869,
+      "loss": 0.7364,
+      "step": 3519
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018091015026007242,
+      "loss": 0.758,
+      "step": 3520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018089958647049247,
+      "loss": 0.7942,
+      "step": 3521
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018088902006747044,
+      "loss": 0.6805,
+      "step": 3522
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018087845105134765,
+      "loss": 0.6613,
+      "step": 3523
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018086787942246557,
+      "loss": 0.7344,
+      "step": 3524
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001808573051811657,
+      "loss": 0.7428,
+      "step": 3525
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018084672832778957,
+      "loss": 0.725,
+      "step": 3526
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018083614886267897,
+      "loss": 0.8221,
+      "step": 3527
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018082556678617563,
+      "loss": 0.8102,
+      "step": 3528
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018081498209862132,
+      "loss": 0.7951,
+      "step": 3529
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001808043948003581,
+      "loss": 0.7079,
+      "step": 3530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018079380489172792,
+      "loss": 0.7847,
+      "step": 3531
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018078321237307288,
+      "loss": 0.7505,
+      "step": 3532
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018077261724473522,
+      "loss": 0.7646,
+      "step": 3533
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018076201950705719,
+      "loss": 0.7178,
+      "step": 3534
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001807514191603811,
+      "loss": 0.7532,
+      "step": 3535
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018074081620504946,
+      "loss": 0.6998,
+      "step": 3536
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018073021064140475,
+      "loss": 0.7464,
+      "step": 3537
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001807196024697896,
+      "loss": 0.7193,
+      "step": 3538
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001807089916905467,
+      "loss": 0.7355,
+      "step": 3539
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018069837830401886,
+      "loss": 0.7741,
+      "step": 3540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018068776231054887,
+      "loss": 0.799,
+      "step": 3541
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018067714371047973,
+      "loss": 0.7243,
+      "step": 3542
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018066652250415448,
+      "loss": 0.6951,
+      "step": 3543
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018065589869191618,
+      "loss": 0.7437,
+      "step": 3544
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001806452722741081,
+      "loss": 0.7513,
+      "step": 3545
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001806346432510735,
+      "loss": 0.7473,
+      "step": 3546
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018062401162315572,
+      "loss": 0.7408,
+      "step": 3547
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018061337739069821,
+      "loss": 0.7113,
+      "step": 3548
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018060274055404455,
+      "loss": 0.7203,
+      "step": 3549
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018059210111353834,
+      "loss": 0.7334,
+      "step": 3550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001805814590695233,
+      "loss": 0.7059,
+      "step": 3551
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018057081442234316,
+      "loss": 0.7218,
+      "step": 3552
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018056016717234185,
+      "loss": 0.7179,
+      "step": 3553
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018054951731986333,
+      "loss": 0.7841,
+      "step": 3554
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001805388648652516,
+      "loss": 0.7794,
+      "step": 3555
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018052820980885082,
+      "loss": 0.6963,
+      "step": 3556
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001805175521510052,
+      "loss": 0.7094,
+      "step": 3557
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018050689189205897,
+      "loss": 0.7755,
+      "step": 3558
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001804962290323566,
+      "loss": 0.7897,
+      "step": 3559
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018048556357224248,
+      "loss": 0.7673,
+      "step": 3560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001804748955120612,
+      "loss": 0.7548,
+      "step": 3561
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018046422485215737,
+      "loss": 0.6875,
+      "step": 3562
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00018045355159287572,
+      "loss": 0.8002,
+      "step": 3563
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.000180442875734561,
+      "loss": 0.7297,
+      "step": 3564
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018043219727755812,
+      "loss": 0.697,
+      "step": 3565
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018042151622221208,
+      "loss": 0.7353,
+      "step": 3566
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001804108325688679,
+      "loss": 0.656,
+      "step": 3567
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001804001463178707,
+      "loss": 0.6711,
+      "step": 3568
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001803894574695657,
+      "loss": 0.6732,
+      "step": 3569
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001803787660242982,
+      "loss": 0.6818,
+      "step": 3570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018036807198241358,
+      "loss": 0.7647,
+      "step": 3571
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018035737534425735,
+      "loss": 0.7589,
+      "step": 3572
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018034667611017505,
+      "loss": 0.7687,
+      "step": 3573
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018033597428051225,
+      "loss": 0.7568,
+      "step": 3574
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018032526985561475,
+      "loss": 0.7225,
+      "step": 3575
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001803145628358283,
+      "loss": 0.7723,
+      "step": 3576
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018030385322149886,
+      "loss": 0.7768,
+      "step": 3577
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001802931410129723,
+      "loss": 0.7663,
+      "step": 3578
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018028242621059474,
+      "loss": 0.7111,
+      "step": 3579
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018027170881471232,
+      "loss": 0.8046,
+      "step": 3580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018026098882567126,
+      "loss": 0.683,
+      "step": 3581
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018025026624381784,
+      "loss": 0.6661,
+      "step": 3582
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018023954106949848,
+      "loss": 0.7438,
+      "step": 3583
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018022881330305963,
+      "loss": 0.7585,
+      "step": 3584
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018021808294484785,
+      "loss": 0.7903,
+      "step": 3585
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001802073499952098,
+      "loss": 0.6978,
+      "step": 3586
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001801966144544922,
+      "loss": 0.7808,
+      "step": 3587
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018018587632304188,
+      "loss": 0.7301,
+      "step": 3588
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018017513560120567,
+      "loss": 0.7108,
+      "step": 3589
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001801643922893306,
+      "loss": 0.712,
+      "step": 3590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018015364638776372,
+      "loss": 0.7057,
+      "step": 3591
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018014289789685214,
+      "loss": 0.7012,
+      "step": 3592
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018013214681694315,
+      "loss": 0.7028,
+      "step": 3593
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018012139314838404,
+      "loss": 0.7099,
+      "step": 3594
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018011063689152214,
+      "loss": 0.7467,
+      "step": 3595
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018009987804670503,
+      "loss": 0.6943,
+      "step": 3596
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018008911661428022,
+      "loss": 0.7254,
+      "step": 3597
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001800783525945953,
+      "loss": 0.7677,
+      "step": 3598
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001800675859879981,
+      "loss": 0.7248,
+      "step": 3599
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001800568167948364,
+      "loss": 0.7081,
+      "step": 3600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001800460450154581,
+      "loss": 0.7909,
+      "step": 3601
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018003527065021114,
+      "loss": 0.7852,
+      "step": 3602
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018002449369944361,
+      "loss": 0.7678,
+      "step": 3603
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018001371416350364,
+      "loss": 0.8182,
+      "step": 3604
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00018000293204273953,
+      "loss": 0.787,
+      "step": 3605
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017999214733749948,
+      "loss": 0.7015,
+      "step": 3606
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017998136004813197,
+      "loss": 0.757,
+      "step": 3607
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017997057017498546,
+      "loss": 0.7423,
+      "step": 3608
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017995977771840852,
+      "loss": 0.7992,
+      "step": 3609
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001799489826787498,
+      "loss": 0.773,
+      "step": 3610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.000179938185056358,
+      "loss": 0.7713,
+      "step": 3611
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017992738485158197,
+      "loss": 0.7985,
+      "step": 3612
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017991658206477059,
+      "loss": 0.783,
+      "step": 3613
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017990577669627284,
+      "loss": 0.6982,
+      "step": 3614
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017989496874643777,
+      "loss": 0.7021,
+      "step": 3615
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001798841582156146,
+      "loss": 0.7493,
+      "step": 3616
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001798733451041525,
+      "loss": 0.7708,
+      "step": 3617
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017986252941240078,
+      "loss": 0.7327,
+      "step": 3618
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017985171114070884,
+      "loss": 0.6444,
+      "step": 3619
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001798408902894262,
+      "loss": 0.7222,
+      "step": 3620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001798300668589024,
+      "loss": 0.7594,
+      "step": 3621
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017981924084948706,
+      "loss": 0.7058,
+      "step": 3622
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017980841226153,
+      "loss": 0.7543,
+      "step": 3623
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017979758109538095,
+      "loss": 0.6835,
+      "step": 3624
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017978674735138982,
+      "loss": 0.745,
+      "step": 3625
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017977591102990662,
+      "loss": 0.8116,
+      "step": 3626
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001797650721312814,
+      "loss": 0.73,
+      "step": 3627
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001797542306558643,
+      "loss": 0.8038,
+      "step": 3628
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017974338660400557,
+      "loss": 0.7233,
+      "step": 3629
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017973253997605548,
+      "loss": 0.7724,
+      "step": 3630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001797216907723645,
+      "loss": 0.7178,
+      "step": 3631
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017971083899328308,
+      "loss": 0.7266,
+      "step": 3632
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017969998463916172,
+      "loss": 0.7259,
+      "step": 3633
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017968912771035119,
+      "loss": 0.7512,
+      "step": 3634
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017967826820720214,
+      "loss": 0.686,
+      "step": 3635
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001796674061300654,
+      "loss": 0.6915,
+      "step": 3636
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017965654147929184,
+      "loss": 0.6945,
+      "step": 3637
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001796456742552325,
+      "loss": 0.6924,
+      "step": 3638
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017963480445823838,
+      "loss": 0.7334,
+      "step": 3639
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017962393208866066,
+      "loss": 0.7315,
+      "step": 3640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017961305714685058,
+      "loss": 0.7139,
+      "step": 3641
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017960217963315943,
+      "loss": 0.6965,
+      "step": 3642
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017959129954793861,
+      "loss": 0.6981,
+      "step": 3643
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001795804168915396,
+      "loss": 0.7111,
+      "step": 3644
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017956953166431398,
+      "loss": 0.7253,
+      "step": 3645
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001795586438666134,
+      "loss": 0.7826,
+      "step": 3646
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001795477534987895,
+      "loss": 0.7554,
+      "step": 3647
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001795368605611942,
+      "loss": 0.7539,
+      "step": 3648
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017952596505417932,
+      "loss": 0.7884,
+      "step": 3649
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017951506697809692,
+      "loss": 0.781,
+      "step": 3650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017950416633329895,
+      "loss": 0.7525,
+      "step": 3651
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017949326312013763,
+      "loss": 0.7307,
+      "step": 3652
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017948235733896515,
+      "loss": 0.668,
+      "step": 3653
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017947144899013387,
+      "loss": 0.793,
+      "step": 3654
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017946053807399613,
+      "loss": 0.7317,
+      "step": 3655
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001794496245909044,
+      "loss": 0.7367,
+      "step": 3656
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017943870854121124,
+      "loss": 0.7777,
+      "step": 3657
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017942778992526936,
+      "loss": 0.6818,
+      "step": 3658
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017941686874343142,
+      "loss": 0.7587,
+      "step": 3659
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017940594499605024,
+      "loss": 0.7402,
+      "step": 3660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001793950186834787,
+      "loss": 0.7058,
+      "step": 3661
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017938408980606977,
+      "loss": 0.7705,
+      "step": 3662
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001793731583641765,
+      "loss": 0.7211,
+      "step": 3663
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017936222435815206,
+      "loss": 0.7898,
+      "step": 3664
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017935128778834964,
+      "loss": 0.7251,
+      "step": 3665
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017934034865512257,
+      "loss": 0.7601,
+      "step": 3666
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017932940695882424,
+      "loss": 0.7283,
+      "step": 3667
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017931846269980805,
+      "loss": 0.7239,
+      "step": 3668
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017930751587842765,
+      "loss": 0.6966,
+      "step": 3669
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017929656649503658,
+      "loss": 0.7066,
+      "step": 3670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017928561454998865,
+      "loss": 0.695,
+      "step": 3671
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017927466004363758,
+      "loss": 0.7535,
+      "step": 3672
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017926370297633728,
+      "loss": 0.6617,
+      "step": 3673
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017925274334844174,
+      "loss": 0.7707,
+      "step": 3674
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.000179241781160305,
+      "loss": 0.7573,
+      "step": 3675
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017923081641228115,
+      "loss": 0.7896,
+      "step": 3676
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017921984910472446,
+      "loss": 0.7003,
+      "step": 3677
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017920887923798917,
+      "loss": 0.8472,
+      "step": 3678
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017919790681242975,
+      "loss": 0.7435,
+      "step": 3679
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017918693182840054,
+      "loss": 0.688,
+      "step": 3680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017917595428625614,
+      "loss": 0.7232,
+      "step": 3681
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017916497418635124,
+      "loss": 0.7513,
+      "step": 3682
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017915399152904047,
+      "loss": 0.7841,
+      "step": 3683
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017914300631467861,
+      "loss": 0.6829,
+      "step": 3684
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017913201854362058,
+      "loss": 0.6923,
+      "step": 3685
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017912102821622137,
+      "loss": 0.7256,
+      "step": 3686
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017911003533283593,
+      "loss": 0.7268,
+      "step": 3687
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017909903989381945,
+      "loss": 0.7311,
+      "step": 3688
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001790880418995271,
+      "loss": 0.7405,
+      "step": 3689
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001790770413503142,
+      "loss": 0.7722,
+      "step": 3690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017906603824653606,
+      "loss": 0.6734,
+      "step": 3691
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017905503258854822,
+      "loss": 0.694,
+      "step": 3692
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017904402437670614,
+      "loss": 0.799,
+      "step": 3693
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001790330136113655,
+      "loss": 0.7622,
+      "step": 3694
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017902200029288194,
+      "loss": 0.717,
+      "step": 3695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001790109844216113,
+      "loss": 0.6928,
+      "step": 3696
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017899996599790937,
+      "loss": 0.7552,
+      "step": 3697
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017898894502213218,
+      "loss": 0.7671,
+      "step": 3698
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001789779214946357,
+      "loss": 0.705,
+      "step": 3699
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017896689541577608,
+      "loss": 0.7286,
+      "step": 3700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017895586678590947,
+      "loss": 0.772,
+      "step": 3701
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017894483560539223,
+      "loss": 0.689,
+      "step": 3702
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017893380187458061,
+      "loss": 0.6885,
+      "step": 3703
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017892276559383112,
+      "loss": 0.7092,
+      "step": 3704
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001789117267635003,
+      "loss": 0.7138,
+      "step": 3705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017890068538394473,
+      "loss": 0.6859,
+      "step": 3706
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017888964145552108,
+      "loss": 0.7367,
+      "step": 3707
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017887859497858614,
+      "loss": 0.7119,
+      "step": 3708
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001788675459534968,
+      "loss": 0.6952,
+      "step": 3709
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001788564943806099,
+      "loss": 0.7262,
+      "step": 3710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017884544026028256,
+      "loss": 0.7183,
+      "step": 3711
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017883438359287182,
+      "loss": 0.7528,
+      "step": 3712
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001788233243787349,
+      "loss": 0.7477,
+      "step": 3713
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017881226261822907,
+      "loss": 0.7191,
+      "step": 3714
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017880119831171161,
+      "loss": 0.7716,
+      "step": 3715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017879013145954002,
+      "loss": 0.7044,
+      "step": 3716
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017877906206207177,
+      "loss": 0.6626,
+      "step": 3717
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017876799011966448,
+      "loss": 0.7173,
+      "step": 3718
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017875691563267582,
+      "loss": 0.767,
+      "step": 3719
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017874583860146356,
+      "loss": 0.7557,
+      "step": 3720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017873475902638553,
+      "loss": 0.749,
+      "step": 3721
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017872367690779966,
+      "loss": 0.7584,
+      "step": 3722
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001787125922460639,
+      "loss": 0.7942,
+      "step": 3723
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017870150504153642,
+      "loss": 0.7938,
+      "step": 3724
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017869041529457533,
+      "loss": 0.7827,
+      "step": 3725
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017867932300553893,
+      "loss": 0.757,
+      "step": 3726
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017866822817478553,
+      "loss": 0.7284,
+      "step": 3727
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017865713080267352,
+      "loss": 0.7686,
+      "step": 3728
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017864603088956144,
+      "loss": 0.7712,
+      "step": 3729
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017863492843580786,
+      "loss": 0.7672,
+      "step": 3730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001786238234417714,
+      "loss": 0.6538,
+      "step": 3731
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017861271590781085,
+      "loss": 0.8068,
+      "step": 3732
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017860160583428508,
+      "loss": 0.7607,
+      "step": 3733
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017859049322155288,
+      "loss": 0.7582,
+      "step": 3734
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017857937806997333,
+      "loss": 0.6521,
+      "step": 3735
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017856826037990545,
+      "loss": 0.7514,
+      "step": 3736
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017855714015170847,
+      "loss": 0.774,
+      "step": 3737
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017854601738574152,
+      "loss": 0.6915,
+      "step": 3738
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.000178534892082364,
+      "loss": 0.7392,
+      "step": 3739
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017852376424193532,
+      "loss": 0.7108,
+      "step": 3740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017851263386481489,
+      "loss": 0.6994,
+      "step": 3741
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001785015009513623,
+      "loss": 0.6499,
+      "step": 3742
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017849036550193726,
+      "loss": 0.7151,
+      "step": 3743
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017847922751689941,
+      "loss": 0.7085,
+      "step": 3744
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017846808699660858,
+      "loss": 0.7268,
+      "step": 3745
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001784569439414247,
+      "loss": 0.7236,
+      "step": 3746
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017844579835170774,
+      "loss": 0.674,
+      "step": 3747
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001784346502278177,
+      "loss": 0.7083,
+      "step": 3748
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001784234995701148,
+      "loss": 0.725,
+      "step": 3749
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017841234637895918,
+      "loss": 0.684,
+      "step": 3750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017840119065471116,
+      "loss": 0.7707,
+      "step": 3751
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017839003239773118,
+      "loss": 0.7248,
+      "step": 3752
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017837887160837966,
+      "loss": 0.7571,
+      "step": 3753
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001783677082870171,
+      "loss": 0.7593,
+      "step": 3754
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017835654243400425,
+      "loss": 0.7113,
+      "step": 3755
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017834537404970172,
+      "loss": 0.7518,
+      "step": 3756
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017833420313447034,
+      "loss": 0.7504,
+      "step": 3757
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017832302968867096,
+      "loss": 0.8169,
+      "step": 3758
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017831185371266453,
+      "loss": 0.7957,
+      "step": 3759
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017830067520681214,
+      "loss": 0.7519,
+      "step": 3760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017828949417147486,
+      "loss": 0.7513,
+      "step": 3761
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001782783106070139,
+      "loss": 0.7451,
+      "step": 3762
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001782671245137906,
+      "loss": 0.7634,
+      "step": 3763
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017825593589216622,
+      "loss": 0.6833,
+      "step": 3764
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001782447447425023,
+      "loss": 0.7486,
+      "step": 3765
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001782335510651603,
+      "loss": 0.723,
+      "step": 3766
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017822235486050187,
+      "loss": 0.768,
+      "step": 3767
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017821115612888869,
+      "loss": 0.7227,
+      "step": 3768
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017819995487068253,
+      "loss": 0.7722,
+      "step": 3769
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017818875108624522,
+      "loss": 0.7718,
+      "step": 3770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001781775447759387,
+      "loss": 0.735,
+      "step": 3771
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017816633594012506,
+      "loss": 0.7396,
+      "step": 3772
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001781551245791663,
+      "loss": 0.6986,
+      "step": 3773
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001781439106934247,
+      "loss": 0.6933,
+      "step": 3774
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017813269428326243,
+      "loss": 0.7058,
+      "step": 3775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001781214753490419,
+      "loss": 0.7284,
+      "step": 3776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017811025389112545,
+      "loss": 0.7159,
+      "step": 3777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017809902990987571,
+      "loss": 0.7778,
+      "step": 3778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017808780340565513,
+      "loss": 0.7495,
+      "step": 3779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001780765743788265,
+      "loss": 0.7719,
+      "step": 3780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017806534282975253,
+      "loss": 0.7314,
+      "step": 3781
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000178054108758796,
+      "loss": 0.669,
+      "step": 3782
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017804287216631988,
+      "loss": 0.6668,
+      "step": 3783
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017803163305268717,
+      "loss": 0.7088,
+      "step": 3784
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017802039141826093,
+      "loss": 0.6985,
+      "step": 3785
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017800914726340432,
+      "loss": 0.6983,
+      "step": 3786
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017799790058848055,
+      "loss": 0.6732,
+      "step": 3787
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000177986651393853,
+      "loss": 0.7656,
+      "step": 3788
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000177975399679885,
+      "loss": 0.8048,
+      "step": 3789
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017796414544694013,
+      "loss": 0.8274,
+      "step": 3790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001779528886953819,
+      "loss": 0.7322,
+      "step": 3791
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017794162942557393,
+      "loss": 0.8289,
+      "step": 3792
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017793036763788,
+      "loss": 0.6809,
+      "step": 3793
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017791910333266387,
+      "loss": 0.6955,
+      "step": 3794
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017790783651028944,
+      "loss": 0.7131,
+      "step": 3795
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017789656717112076,
+      "loss": 0.7516,
+      "step": 3796
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017788529531552176,
+      "loss": 0.7415,
+      "step": 3797
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017787402094385666,
+      "loss": 0.7337,
+      "step": 3798
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017786274405648966,
+      "loss": 0.7386,
+      "step": 3799
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017785146465378504,
+      "loss": 0.7665,
+      "step": 3800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001778401827361072,
+      "loss": 0.7355,
+      "step": 3801
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017782889830382056,
+      "loss": 0.8128,
+      "step": 3802
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017781761135728973,
+      "loss": 0.7794,
+      "step": 3803
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017780632189687924,
+      "loss": 0.7433,
+      "step": 3804
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017779502992295389,
+      "loss": 0.719,
+      "step": 3805
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017778373543587842,
+      "loss": 0.6649,
+      "step": 3806
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017777243843601764,
+      "loss": 0.7075,
+      "step": 3807
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001777611389237366,
+      "loss": 0.7363,
+      "step": 3808
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017774983689940025,
+      "loss": 0.8306,
+      "step": 3809
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017773853236337375,
+      "loss": 0.776,
+      "step": 3810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017772722531602227,
+      "loss": 0.735,
+      "step": 3811
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017771591575771106,
+      "loss": 0.6973,
+      "step": 3812
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017770460368880553,
+      "loss": 0.7455,
+      "step": 3813
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017769328910967105,
+      "loss": 0.7407,
+      "step": 3814
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017768197202067316,
+      "loss": 0.7329,
+      "step": 3815
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017767065242217745,
+      "loss": 0.6722,
+      "step": 3816
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017765933031454965,
+      "loss": 0.7346,
+      "step": 3817
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001776480056981554,
+      "loss": 0.7142,
+      "step": 3818
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017763667857336065,
+      "loss": 0.6947,
+      "step": 3819
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017762534894053128,
+      "loss": 0.7382,
+      "step": 3820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001776140168000333,
+      "loss": 0.6609,
+      "step": 3821
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001776026821522328,
+      "loss": 0.7452,
+      "step": 3822
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001775913449974959,
+      "loss": 0.7788,
+      "step": 3823
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001775800053361889,
+      "loss": 0.679,
+      "step": 3824
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017756866316867806,
+      "loss": 0.7504,
+      "step": 3825
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017755731849532984,
+      "loss": 0.6656,
+      "step": 3826
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001775459713165107,
+      "loss": 0.7475,
+      "step": 3827
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017753462163258723,
+      "loss": 0.682,
+      "step": 3828
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001775232694439261,
+      "loss": 0.7403,
+      "step": 3829
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017751191475089394,
+      "loss": 0.7283,
+      "step": 3830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017750055755385766,
+      "loss": 0.7642,
+      "step": 3831
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001774891978531841,
+      "loss": 0.6881,
+      "step": 3832
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001774778356492403,
+      "loss": 0.7064,
+      "step": 3833
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017746647094239325,
+      "loss": 0.8064,
+      "step": 3834
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017745510373301006,
+      "loss": 0.7206,
+      "step": 3835
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017744373402145804,
+      "loss": 0.7644,
+      "step": 3836
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017743236180810438,
+      "loss": 0.7704,
+      "step": 3837
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017742098709331653,
+      "loss": 0.751,
+      "step": 3838
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017740960987746194,
+      "loss": 0.7394,
+      "step": 3839
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001773982301609081,
+      "loss": 0.7245,
+      "step": 3840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017738684794402267,
+      "loss": 0.7367,
+      "step": 3841
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017737546322717335,
+      "loss": 0.6991,
+      "step": 3842
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017736407601072788,
+      "loss": 0.7892,
+      "step": 3843
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001773526862950542,
+      "loss": 0.7129,
+      "step": 3844
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017734129408052018,
+      "loss": 0.8246,
+      "step": 3845
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017732989936749389,
+      "loss": 0.7357,
+      "step": 3846
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017731850215634337,
+      "loss": 0.6763,
+      "step": 3847
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001773071024474369,
+      "loss": 0.7008,
+      "step": 3848
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017729570024114266,
+      "loss": 0.7174,
+      "step": 3849
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.000177284295537829,
+      "loss": 0.7364,
+      "step": 3850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017727288833786442,
+      "loss": 0.7526,
+      "step": 3851
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017726147864161734,
+      "loss": 0.7508,
+      "step": 3852
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017725006644945642,
+      "loss": 0.7128,
+      "step": 3853
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001772386517617503,
+      "loss": 0.7338,
+      "step": 3854
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017722723457886769,
+      "loss": 0.7564,
+      "step": 3855
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017721581490117748,
+      "loss": 0.7493,
+      "step": 3856
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017720439272904853,
+      "loss": 0.8075,
+      "step": 3857
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001771929680628499,
+      "loss": 0.7023,
+      "step": 3858
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017718154090295057,
+      "loss": 0.7542,
+      "step": 3859
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017717011124971974,
+      "loss": 0.6887,
+      "step": 3860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017715867910352663,
+      "loss": 0.7362,
+      "step": 3861
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001771472444647406,
+      "loss": 0.6912,
+      "step": 3862
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017713580733373099,
+      "loss": 0.7312,
+      "step": 3863
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017712436771086728,
+      "loss": 0.7056,
+      "step": 3864
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017711292559651905,
+      "loss": 0.7298,
+      "step": 3865
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017710148099105591,
+      "loss": 0.7727,
+      "step": 3866
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017709003389484757,
+      "loss": 0.7615,
+      "step": 3867
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00017707858430826386,
+      "loss": 0.7524,
+      "step": 3868
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001770671322316746,
+      "loss": 0.73,
+      "step": 3869
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017705567766544982,
+      "loss": 0.7491,
+      "step": 3870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001770442206099595,
+      "loss": 0.7679,
+      "step": 3871
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001770327610655738,
+      "loss": 0.7146,
+      "step": 3872
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017702129903266286,
+      "loss": 0.6603,
+      "step": 3873
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017700983451159703,
+      "loss": 0.7056,
+      "step": 3874
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017699836750274662,
+      "loss": 0.7145,
+      "step": 3875
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017698689800648207,
+      "loss": 0.7159,
+      "step": 3876
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017697542602317388,
+      "loss": 0.7474,
+      "step": 3877
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017696395155319272,
+      "loss": 0.7577,
+      "step": 3878
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017695247459690924,
+      "loss": 0.6663,
+      "step": 3879
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001769409951546942,
+      "loss": 0.734,
+      "step": 3880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001769295132269184,
+      "loss": 0.7741,
+      "step": 3881
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017691802881395277,
+      "loss": 0.7025,
+      "step": 3882
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017690654191616836,
+      "loss": 0.7989,
+      "step": 3883
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001768950525339362,
+      "loss": 0.7454,
+      "step": 3884
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017688356066762753,
+      "loss": 0.6688,
+      "step": 3885
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001768720663176135,
+      "loss": 0.764,
+      "step": 3886
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017686056948426545,
+      "loss": 0.8,
+      "step": 3887
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017684907016795483,
+      "loss": 0.6924,
+      "step": 3888
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017683756836905313,
+      "loss": 0.7146,
+      "step": 3889
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017682606408793184,
+      "loss": 0.7151,
+      "step": 3890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017681455732496262,
+      "loss": 0.6748,
+      "step": 3891
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017680304808051728,
+      "loss": 0.7575,
+      "step": 3892
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017679153635496747,
+      "loss": 0.7248,
+      "step": 3893
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017678002214868523,
+      "loss": 0.6924,
+      "step": 3894
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017676850546204242,
+      "loss": 0.7151,
+      "step": 3895
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017675698629541112,
+      "loss": 0.7446,
+      "step": 3896
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017674546464916348,
+      "loss": 0.7484,
+      "step": 3897
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017673394052367168,
+      "loss": 0.637,
+      "step": 3898
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.000176722413919308,
+      "loss": 0.734,
+      "step": 3899
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001767108848364448,
+      "loss": 0.7003,
+      "step": 3900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017669935327545451,
+      "loss": 0.8089,
+      "step": 3901
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001766878192367097,
+      "loss": 0.6805,
+      "step": 3902
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001766762827205829,
+      "loss": 0.6957,
+      "step": 3903
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001766647437274469,
+      "loss": 0.7021,
+      "step": 3904
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001766532022576744,
+      "loss": 0.6858,
+      "step": 3905
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017664165831163823,
+      "loss": 0.8213,
+      "step": 3906
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017663011188971136,
+      "loss": 0.7186,
+      "step": 3907
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017661856299226676,
+      "loss": 0.6675,
+      "step": 3908
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017660701161967754,
+      "loss": 0.7519,
+      "step": 3909
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017659545777231684,
+      "loss": 0.7177,
+      "step": 3910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001765839014505579,
+      "loss": 0.686,
+      "step": 3911
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017657234265477408,
+      "loss": 0.7419,
+      "step": 3912
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017656078138533873,
+      "loss": 0.6943,
+      "step": 3913
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017654921764262538,
+      "loss": 0.7431,
+      "step": 3914
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001765376514270076,
+      "loss": 0.702,
+      "step": 3915
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017652608273885901,
+      "loss": 0.7544,
+      "step": 3916
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017651451157855332,
+      "loss": 0.7461,
+      "step": 3917
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001765029379464644,
+      "loss": 0.6693,
+      "step": 3918
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017649136184296605,
+      "loss": 0.6945,
+      "step": 3919
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017647978326843227,
+      "loss": 0.7955,
+      "step": 3920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001764682022232371,
+      "loss": 0.7689,
+      "step": 3921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017645661870775465,
+      "loss": 0.752,
+      "step": 3922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017644503272235915,
+      "loss": 0.6039,
+      "step": 3923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001764334442674249,
+      "loss": 0.7791,
+      "step": 3924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017642185334332616,
+      "loss": 0.7153,
+      "step": 3925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001764102599504375,
+      "loss": 0.6949,
+      "step": 3926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001763986640891334,
+      "loss": 0.7442,
+      "step": 3927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017638706575978842,
+      "loss": 0.6922,
+      "step": 3928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017637546496277728,
+      "loss": 0.7488,
+      "step": 3929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017636386169847472,
+      "loss": 0.6723,
+      "step": 3930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001763522559672556,
+      "loss": 0.7208,
+      "step": 3931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017634064776949485,
+      "loss": 0.7223,
+      "step": 3932
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001763290371055674,
+      "loss": 0.7544,
+      "step": 3933
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017631742397584843,
+      "loss": 0.7193,
+      "step": 3934
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000176305808380713,
+      "loss": 0.732,
+      "step": 3935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017629419032053645,
+      "loss": 0.7704,
+      "step": 3936
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017628256979569404,
+      "loss": 0.7051,
+      "step": 3937
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017627094680656118,
+      "loss": 0.7687,
+      "step": 3938
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001762593213535133,
+      "loss": 0.7516,
+      "step": 3939
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017624769343692605,
+      "loss": 0.7462,
+      "step": 3940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000176236063057175,
+      "loss": 0.6954,
+      "step": 3941
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017622443021463591,
+      "loss": 0.7283,
+      "step": 3942
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017621279490968452,
+      "loss": 0.7078,
+      "step": 3943
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017620115714269674,
+      "loss": 0.7875,
+      "step": 3944
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017618951691404856,
+      "loss": 0.7684,
+      "step": 3945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017617787422411594,
+      "loss": 0.6831,
+      "step": 3946
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017616622907327503,
+      "loss": 0.7263,
+      "step": 3947
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017615458146190207,
+      "loss": 0.8775,
+      "step": 3948
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017614293139037326,
+      "loss": 0.6574,
+      "step": 3949
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000176131278859065,
+      "loss": 0.7303,
+      "step": 3950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017611962386835365,
+      "loss": 0.7711,
+      "step": 3951
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017610796641861581,
+      "loss": 0.7671,
+      "step": 3952
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017609630651022805,
+      "loss": 0.7715,
+      "step": 3953
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017608464414356703,
+      "loss": 0.7869,
+      "step": 3954
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001760729793190095,
+      "loss": 0.7519,
+      "step": 3955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017606131203693227,
+      "loss": 0.7342,
+      "step": 3956
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017604964229771226,
+      "loss": 0.7263,
+      "step": 3957
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001760379701017265,
+      "loss": 0.7007,
+      "step": 3958
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.000176026295449352,
+      "loss": 0.698,
+      "step": 3959
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017601461834096594,
+      "loss": 0.7376,
+      "step": 3960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001760029387769455,
+      "loss": 0.6845,
+      "step": 3961
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017599125675766807,
+      "loss": 0.7662,
+      "step": 3962
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017597957228351093,
+      "loss": 0.7474,
+      "step": 3963
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017596788535485165,
+      "loss": 0.8055,
+      "step": 3964
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017595619597206768,
+      "loss": 0.8463,
+      "step": 3965
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017594450413553674,
+      "loss": 0.7222,
+      "step": 3966
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017593280984563644,
+      "loss": 0.7108,
+      "step": 3967
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017592111310274462,
+      "loss": 0.7884,
+      "step": 3968
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017590941390723913,
+      "loss": 0.7548,
+      "step": 3969
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017589771225949785,
+      "loss": 0.6766,
+      "step": 3970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017588600815989888,
+      "loss": 0.7685,
+      "step": 3971
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001758743016088203,
+      "loss": 0.6425,
+      "step": 3972
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017586259260664026,
+      "loss": 0.7352,
+      "step": 3973
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017585088115373702,
+      "loss": 0.7615,
+      "step": 3974
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017583916725048894,
+      "loss": 0.8306,
+      "step": 3975
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017582745089727442,
+      "loss": 0.7011,
+      "step": 3976
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017581573209447194,
+      "loss": 0.6871,
+      "step": 3977
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017580401084246012,
+      "loss": 0.7234,
+      "step": 3978
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017579228714161759,
+      "loss": 0.6655,
+      "step": 3979
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017578056099232304,
+      "loss": 0.7159,
+      "step": 3980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017576883239495532,
+      "loss": 0.681,
+      "step": 3981
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001757571013498933,
+      "loss": 0.7135,
+      "step": 3982
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017574536785751598,
+      "loss": 0.7086,
+      "step": 3983
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001757336319182024,
+      "loss": 0.7825,
+      "step": 3984
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017572189353233167,
+      "loss": 0.7793,
+      "step": 3985
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017571015270028298,
+      "loss": 0.7367,
+      "step": 3986
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017569840942243566,
+      "loss": 0.7609,
+      "step": 3987
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.000175686663699169,
+      "loss": 0.7121,
+      "step": 3988
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017567491553086256,
+      "loss": 0.7723,
+      "step": 3989
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017566316491789577,
+      "loss": 0.741,
+      "step": 3990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001756514118606482,
+      "loss": 0.7548,
+      "step": 3991
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017563965635949966,
+      "loss": 0.7659,
+      "step": 3992
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017562789841482984,
+      "loss": 0.7693,
+      "step": 3993
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017561613802701852,
+      "loss": 0.7654,
+      "step": 3994
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001756043751964457,
+      "loss": 0.6721,
+      "step": 3995
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017559260992349135,
+      "loss": 0.7191,
+      "step": 3996
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017558084220853552,
+      "loss": 0.728,
+      "step": 3997
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017556907205195835,
+      "loss": 0.7921,
+      "step": 3998
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017555729945414015,
+      "loss": 0.6883,
+      "step": 3999
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017554552441546118,
+      "loss": 0.7344,
+      "step": 4000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017553374693630183,
+      "loss": 0.7185,
+      "step": 4001
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017552196701704255,
+      "loss": 0.7164,
+      "step": 4002
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017551018465806394,
+      "loss": 0.7275,
+      "step": 4003
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017549839985974658,
+      "loss": 0.7314,
+      "step": 4004
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017548661262247118,
+      "loss": 0.7452,
+      "step": 4005
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001754748229466186,
+      "loss": 0.7259,
+      "step": 4006
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001754630308325696,
+      "loss": 0.6833,
+      "step": 4007
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017545123628070512,
+      "loss": 0.7711,
+      "step": 4008
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017543943929140626,
+      "loss": 0.7712,
+      "step": 4009
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017542763986505406,
+      "loss": 0.7305,
+      "step": 4010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017541583800202975,
+      "loss": 0.6528,
+      "step": 4011
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017540403370271452,
+      "loss": 0.7301,
+      "step": 4012
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017539222696748974,
+      "loss": 0.7674,
+      "step": 4013
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017538041779673684,
+      "loss": 0.6842,
+      "step": 4014
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001753686061908373,
+      "loss": 0.7278,
+      "step": 4015
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017535679215017265,
+      "loss": 0.7899,
+      "step": 4016
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001753449756751246,
+      "loss": 0.6954,
+      "step": 4017
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017533315676607485,
+      "loss": 0.7284,
+      "step": 4018
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001753213354234052,
+      "loss": 0.741,
+      "step": 4019
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017530951164749757,
+      "loss": 0.7111,
+      "step": 4020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017529768543873391,
+      "loss": 0.7609,
+      "step": 4021
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017528585679749622,
+      "loss": 0.6965,
+      "step": 4022
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001752740257241667,
+      "loss": 0.7715,
+      "step": 4023
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017526219221912745,
+      "loss": 0.751,
+      "step": 4024
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017525035628276084,
+      "loss": 0.7118,
+      "step": 4025
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017523851791544917,
+      "loss": 0.7886,
+      "step": 4026
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017522667711757493,
+      "loss": 0.7344,
+      "step": 4027
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017521483388952057,
+      "loss": 0.7038,
+      "step": 4028
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017520298823166873,
+      "loss": 0.709,
+      "step": 4029
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017519114014440205,
+      "loss": 0.6957,
+      "step": 4030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017517928962810332,
+      "loss": 0.7512,
+      "step": 4031
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017516743668315532,
+      "loss": 0.6972,
+      "step": 4032
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.000175155581309941,
+      "loss": 0.7655,
+      "step": 4033
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001751437235088433,
+      "loss": 0.7971,
+      "step": 4034
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001751318632802453,
+      "loss": 0.6921,
+      "step": 4035
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017512000062453018,
+      "loss": 0.6728,
+      "step": 4036
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017510813554208113,
+      "loss": 0.7658,
+      "step": 4037
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001750962680332814,
+      "loss": 0.7264,
+      "step": 4038
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017508439809851446,
+      "loss": 0.7455,
+      "step": 4039
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017507252573816372,
+      "loss": 0.8065,
+      "step": 4040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017506065095261272,
+      "loss": 0.7081,
+      "step": 4041
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017504877374224505,
+      "loss": 0.7197,
+      "step": 4042
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00017503689410744444,
+      "loss": 0.7322,
+      "step": 4043
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001750250120485946,
+      "loss": 0.7423,
+      "step": 4044
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017501312756607943,
+      "loss": 0.728,
+      "step": 4045
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017500124066028287,
+      "loss": 0.7003,
+      "step": 4046
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017498935133158884,
+      "loss": 0.73,
+      "step": 4047
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017497745958038152,
+      "loss": 0.7,
+      "step": 4048
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000174965565407045,
+      "loss": 0.7102,
+      "step": 4049
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017495366881196356,
+      "loss": 0.7852,
+      "step": 4050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017494176979552148,
+      "loss": 0.7414,
+      "step": 4051
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001749298683581032,
+      "loss": 0.7295,
+      "step": 4052
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017491796450009315,
+      "loss": 0.7337,
+      "step": 4053
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001749060582218759,
+      "loss": 0.7538,
+      "step": 4054
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017489414952383609,
+      "loss": 0.826,
+      "step": 4055
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001748822384063584,
+      "loss": 0.7729,
+      "step": 4056
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001748703248698276,
+      "loss": 0.7017,
+      "step": 4057
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017485840891462865,
+      "loss": 0.7355,
+      "step": 4058
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017484649054114643,
+      "loss": 0.6612,
+      "step": 4059
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017483456974976593,
+      "loss": 0.7035,
+      "step": 4060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017482264654087226,
+      "loss": 0.6407,
+      "step": 4061
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017481072091485067,
+      "loss": 0.7562,
+      "step": 4062
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001747987928720863,
+      "loss": 0.7414,
+      "step": 4063
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017478686241296457,
+      "loss": 0.7337,
+      "step": 4064
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017477492953787087,
+      "loss": 0.7392,
+      "step": 4065
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017476299424719067,
+      "loss": 0.6236,
+      "step": 4066
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001747510565413095,
+      "loss": 0.7613,
+      "step": 4067
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017473911642061315,
+      "loss": 0.7173,
+      "step": 4068
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017472717388548715,
+      "loss": 0.7208,
+      "step": 4069
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017471522893631748,
+      "loss": 0.7767,
+      "step": 4070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017470328157348988,
+      "loss": 0.739,
+      "step": 4071
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017469133179739036,
+      "loss": 0.6726,
+      "step": 4072
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017467937960840497,
+      "loss": 0.7535,
+      "step": 4073
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017466742500691978,
+      "loss": 0.7182,
+      "step": 4074
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017465546799332103,
+      "loss": 0.7264,
+      "step": 4075
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000174643508567995,
+      "loss": 0.7235,
+      "step": 4076
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017463154673132798,
+      "loss": 0.7466,
+      "step": 4077
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001746195824837064,
+      "loss": 0.8107,
+      "step": 4078
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017460761582551682,
+      "loss": 0.7232,
+      "step": 4079
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017459564675714571,
+      "loss": 0.716,
+      "step": 4080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001745836752789799,
+      "loss": 0.7216,
+      "step": 4081
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017457170139140594,
+      "loss": 0.7476,
+      "step": 4082
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017455972509481077,
+      "loss": 0.7228,
+      "step": 4083
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001745477463895812,
+      "loss": 0.7338,
+      "step": 4084
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017453576527610428,
+      "loss": 0.6972,
+      "step": 4085
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000174523781754767,
+      "loss": 0.669,
+      "step": 4086
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017451179582595648,
+      "loss": 0.7375,
+      "step": 4087
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017449980749005998,
+      "loss": 0.7266,
+      "step": 4088
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017448781674746472,
+      "loss": 0.7005,
+      "step": 4089
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017447582359855809,
+      "loss": 0.7218,
+      "step": 4090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001744638280437275,
+      "loss": 0.7548,
+      "step": 4091
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001744518300833605,
+      "loss": 0.8237,
+      "step": 4092
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017443982971784464,
+      "loss": 0.7208,
+      "step": 4093
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017442782694756764,
+      "loss": 0.6925,
+      "step": 4094
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001744158217729172,
+      "loss": 0.7451,
+      "step": 4095
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017440381419428115,
+      "loss": 0.755,
+      "step": 4096
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017439180421204744,
+      "loss": 0.6942,
+      "step": 4097
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.000174379791826604,
+      "loss": 0.7047,
+      "step": 4098
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001743677770383389,
+      "loss": 0.7973,
+      "step": 4099
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017435575984764027,
+      "loss": 0.7234,
+      "step": 4100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017434374025489632,
+      "loss": 0.7643,
+      "step": 4101
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017433171826049536,
+      "loss": 0.7115,
+      "step": 4102
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017431969386482577,
+      "loss": 0.7263,
+      "step": 4103
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001743076670682759,
+      "loss": 0.7732,
+      "step": 4104
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001742956378712344,
+      "loss": 0.7753,
+      "step": 4105
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017428360627408978,
+      "loss": 0.7706,
+      "step": 4106
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001742715722772308,
+      "loss": 0.7376,
+      "step": 4107
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017425953588104615,
+      "loss": 0.723,
+      "step": 4108
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017424749708592468,
+      "loss": 0.7641,
+      "step": 4109
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001742354558922553,
+      "loss": 0.7822,
+      "step": 4110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.000174223412300427,
+      "loss": 0.6975,
+      "step": 4111
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017421136631082885,
+      "loss": 0.7476,
+      "step": 4112
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017419931792384997,
+      "loss": 0.7668,
+      "step": 4113
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001741872671398796,
+      "loss": 0.7758,
+      "step": 4114
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017417521395930705,
+      "loss": 0.7364,
+      "step": 4115
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017416315838252166,
+      "loss": 0.6875,
+      "step": 4116
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017415110040991294,
+      "loss": 0.7166,
+      "step": 4117
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017413904004187037,
+      "loss": 0.6916,
+      "step": 4118
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017412697727878357,
+      "loss": 0.7466,
+      "step": 4119
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001741149121210422,
+      "loss": 0.7578,
+      "step": 4120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017410284456903608,
+      "loss": 0.7164,
+      "step": 4121
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.000174090774623155,
+      "loss": 0.7327,
+      "step": 4122
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017407870228378888,
+      "loss": 0.7957,
+      "step": 4123
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017406662755132775,
+      "loss": 0.743,
+      "step": 4124
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017405455042616166,
+      "loss": 0.6354,
+      "step": 4125
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017404247090868077,
+      "loss": 0.685,
+      "step": 4126
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017403038899927524,
+      "loss": 0.7357,
+      "step": 4127
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017401830469833548,
+      "loss": 0.801,
+      "step": 4128
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001740062180062518,
+      "loss": 0.715,
+      "step": 4129
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017399412892341466,
+      "loss": 0.7848,
+      "step": 4130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017398203745021463,
+      "loss": 0.7351,
+      "step": 4131
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001739699435870423,
+      "loss": 0.7116,
+      "step": 4132
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017395784733428835,
+      "loss": 0.784,
+      "step": 4133
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017394574869234356,
+      "loss": 0.7337,
+      "step": 4134
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017393364766159878,
+      "loss": 0.7563,
+      "step": 4135
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017392154424244492,
+      "loss": 0.7306,
+      "step": 4136
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017390943843527298,
+      "loss": 0.6884,
+      "step": 4137
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017389733024047405,
+      "loss": 0.8069,
+      "step": 4138
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017388521965843928,
+      "loss": 0.7024,
+      "step": 4139
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017387310668955986,
+      "loss": 0.6874,
+      "step": 4140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017386099133422713,
+      "loss": 0.7097,
+      "step": 4141
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017384887359283248,
+      "loss": 0.6886,
+      "step": 4142
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017383675346576737,
+      "loss": 0.7251,
+      "step": 4143
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017382463095342326,
+      "loss": 0.7911,
+      "step": 4144
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001738125060561919,
+      "loss": 0.7226,
+      "step": 4145
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017380037877446491,
+      "loss": 0.7444,
+      "step": 4146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017378824910863405,
+      "loss": 0.7252,
+      "step": 4147
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017377611705909115,
+      "loss": 0.7435,
+      "step": 4148
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001737639826262282,
+      "loss": 0.7423,
+      "step": 4149
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017375184581043716,
+      "loss": 0.7012,
+      "step": 4150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017373970661211012,
+      "loss": 0.7026,
+      "step": 4151
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001737275650316392,
+      "loss": 0.7648,
+      "step": 4152
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017371542106941665,
+      "loss": 0.7238,
+      "step": 4153
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001737032747258348,
+      "loss": 0.7709,
+      "step": 4154
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.000173691126001286,
+      "loss": 0.7218,
+      "step": 4155
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017367897489616274,
+      "loss": 0.7003,
+      "step": 4156
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017366682141085752,
+      "loss": 0.7725,
+      "step": 4157
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.000173654665545763,
+      "loss": 0.7741,
+      "step": 4158
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017364250730127187,
+      "loss": 0.6379,
+      "step": 4159
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001736303466777769,
+      "loss": 0.6785,
+      "step": 4160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001736181836756709,
+      "loss": 0.7495,
+      "step": 4161
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017360601829534677,
+      "loss": 0.7408,
+      "step": 4162
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001735938505371976,
+      "loss": 0.6767,
+      "step": 4163
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017358168040161641,
+      "loss": 0.6891,
+      "step": 4164
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017356950788899637,
+      "loss": 0.7254,
+      "step": 4165
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001735573329997307,
+      "loss": 0.7718,
+      "step": 4166
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001735451557342127,
+      "loss": 0.7264,
+      "step": 4167
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017353297609283577,
+      "loss": 0.7522,
+      "step": 4168
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017352079407599336,
+      "loss": 0.6659,
+      "step": 4169
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.000173508609684079,
+      "loss": 0.7429,
+      "step": 4170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017349642291748633,
+      "loss": 0.755,
+      "step": 4171
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017348423377660903,
+      "loss": 0.7335,
+      "step": 4172
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017347204226184087,
+      "loss": 0.7203,
+      "step": 4173
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017345984837357565,
+      "loss": 0.7473,
+      "step": 4174
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017344765211220735,
+      "loss": 0.7795,
+      "step": 4175
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017343545347812996,
+      "loss": 0.729,
+      "step": 4176
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017342325247173752,
+      "loss": 0.7839,
+      "step": 4177
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001734110490934242,
+      "loss": 0.7395,
+      "step": 4178
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017339884334358424,
+      "loss": 0.7116,
+      "step": 4179
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017338663522261188,
+      "loss": 0.7538,
+      "step": 4180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001733744247309016,
+      "loss": 0.7584,
+      "step": 4181
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017336221186884784,
+      "loss": 0.7267,
+      "step": 4182
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017334999663684504,
+      "loss": 0.752,
+      "step": 4183
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001733377790352879,
+      "loss": 0.7166,
+      "step": 4184
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017332555906457108,
+      "loss": 0.6902,
+      "step": 4185
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017331333672508933,
+      "loss": 0.7194,
+      "step": 4186
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001733011120172375,
+      "loss": 0.712,
+      "step": 4187
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017328888494141053,
+      "loss": 0.753,
+      "step": 4188
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017327665549800338,
+      "loss": 0.7243,
+      "step": 4189
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017326442368741112,
+      "loss": 0.7976,
+      "step": 4190
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017325218951002894,
+      "loss": 0.7779,
+      "step": 4191
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017323995296625197,
+      "loss": 0.7091,
+      "step": 4192
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017322771405647562,
+      "loss": 0.8119,
+      "step": 4193
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017321547278109517,
+      "loss": 0.6764,
+      "step": 4194
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017320322914050616,
+      "loss": 0.7449,
+      "step": 4195
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017319098313510405,
+      "loss": 0.6812,
+      "step": 4196
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017317873476528446,
+      "loss": 0.7222,
+      "step": 4197
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017316648403144306,
+      "loss": 0.8074,
+      "step": 4198
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017315423093397565,
+      "loss": 0.8048,
+      "step": 4199
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017314197547327803,
+      "loss": 0.7474,
+      "step": 4200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017312971764974614,
+      "loss": 0.7141,
+      "step": 4201
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001731174574637759,
+      "loss": 0.7133,
+      "step": 4202
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017310519491576344,
+      "loss": 0.6757,
+      "step": 4203
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017309293000610486,
+      "loss": 0.7442,
+      "step": 4204
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001730806627351964,
+      "loss": 0.6816,
+      "step": 4205
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017306839310343434,
+      "loss": 0.7223,
+      "step": 4206
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017305612111121504,
+      "loss": 0.7335,
+      "step": 4207
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017304384675893495,
+      "loss": 0.7614,
+      "step": 4208
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017303157004699063,
+      "loss": 0.6839,
+      "step": 4209
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017301929097577858,
+      "loss": 0.6954,
+      "step": 4210
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017300700954569557,
+      "loss": 0.7716,
+      "step": 4211
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001729947257571383,
+      "loss": 0.7431,
+      "step": 4212
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001729824396105036,
+      "loss": 0.7765,
+      "step": 4213
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017297015110618837,
+      "loss": 0.7512,
+      "step": 4214
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001729578602445896,
+      "loss": 0.7505,
+      "step": 4215
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017294556702610432,
+      "loss": 0.7239,
+      "step": 4216
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017293327145112973,
+      "loss": 0.6743,
+      "step": 4217
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017292097352006292,
+      "loss": 0.7004,
+      "step": 4218
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017290867323330127,
+      "loss": 0.7844,
+      "step": 4219
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017289637059124207,
+      "loss": 0.6896,
+      "step": 4220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017288406559428282,
+      "loss": 0.753,
+      "step": 4221
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017287175824282094,
+      "loss": 0.7251,
+      "step": 4222
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001728594485372541,
+      "loss": 0.7586,
+      "step": 4223
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017284713647797997,
+      "loss": 0.6882,
+      "step": 4224
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017283482206539622,
+      "loss": 0.6766,
+      "step": 4225
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017282250529990068,
+      "loss": 0.797,
+      "step": 4226
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001728101861818913,
+      "loss": 0.6972,
+      "step": 4227
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017279786471176593,
+      "loss": 0.7184,
+      "step": 4228
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017278554088992273,
+      "loss": 0.6487,
+      "step": 4229
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001727732147167598,
+      "loss": 0.6427,
+      "step": 4230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017276088619267525,
+      "loss": 0.7455,
+      "step": 4231
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017274855531806743,
+      "loss": 0.7067,
+      "step": 4232
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017273622209333466,
+      "loss": 0.772,
+      "step": 4233
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017272388651887538,
+      "loss": 0.7628,
+      "step": 4234
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017271154859508807,
+      "loss": 0.7918,
+      "step": 4235
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001726992083223713,
+      "loss": 0.7105,
+      "step": 4236
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001726868657011237,
+      "loss": 0.7671,
+      "step": 4237
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017267452073174404,
+      "loss": 0.7254,
+      "step": 4238
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017266217341463112,
+      "loss": 0.726,
+      "step": 4239
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017264982375018377,
+      "loss": 0.7848,
+      "step": 4240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017263747173880097,
+      "loss": 0.7008,
+      "step": 4241
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001726251173808818,
+      "loss": 0.7324,
+      "step": 4242
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017261276067682526,
+      "loss": 0.6856,
+      "step": 4243
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017260040162703062,
+      "loss": 0.7241,
+      "step": 4244
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001725880402318971,
+      "loss": 0.729,
+      "step": 4245
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.000172575676491824,
+      "loss": 0.7511,
+      "step": 4246
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017256331040721082,
+      "loss": 0.6552,
+      "step": 4247
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017255094197845693,
+      "loss": 0.7476,
+      "step": 4248
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017253857120596197,
+      "loss": 0.7671,
+      "step": 4249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017252619809012557,
+      "loss": 0.73,
+      "step": 4250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017251382263134735,
+      "loss": 0.7076,
+      "step": 4251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017250144483002723,
+      "loss": 0.7691,
+      "step": 4252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.000172489064686565,
+      "loss": 0.7275,
+      "step": 4253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001724766822013606,
+      "loss": 0.8123,
+      "step": 4254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017246429737481403,
+      "loss": 0.6975,
+      "step": 4255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017245191020732542,
+      "loss": 0.7161,
+      "step": 4256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001724395206992949,
+      "loss": 0.83,
+      "step": 4257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017242712885112276,
+      "loss": 0.6711,
+      "step": 4258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017241473466320922,
+      "loss": 0.7369,
+      "step": 4259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017240233813595478,
+      "loss": 0.7635,
+      "step": 4260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017238993926975985,
+      "loss": 0.7284,
+      "step": 4261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017237753806502494,
+      "loss": 0.7544,
+      "step": 4262
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017236513452215075,
+      "loss": 0.7629,
+      "step": 4263
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001723527286415379,
+      "loss": 0.7492,
+      "step": 4264
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001723403204235872,
+      "loss": 0.7892,
+      "step": 4265
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001723279098686995,
+      "loss": 0.76,
+      "step": 4266
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001723154969772757,
+      "loss": 0.75,
+      "step": 4267
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017230308174971678,
+      "loss": 0.6667,
+      "step": 4268
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017229066418642384,
+      "loss": 0.7265,
+      "step": 4269
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017227824428779805,
+      "loss": 0.6313,
+      "step": 4270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017226582205424055,
+      "loss": 0.7116,
+      "step": 4271
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001722533974861527,
+      "loss": 0.7723,
+      "step": 4272
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017224097058393589,
+      "loss": 0.7318,
+      "step": 4273
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017222854134799152,
+      "loss": 0.6772,
+      "step": 4274
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017221610977872113,
+      "loss": 0.7073,
+      "step": 4275
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017220367587652627,
+      "loss": 0.7027,
+      "step": 4276
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017219123964180874,
+      "loss": 0.72,
+      "step": 4277
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017217880107497018,
+      "loss": 0.7698,
+      "step": 4278
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017216636017641245,
+      "loss": 0.7144,
+      "step": 4279
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017215391694653747,
+      "loss": 0.7556,
+      "step": 4280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017214147138574717,
+      "loss": 0.7305,
+      "step": 4281
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017212902349444362,
+      "loss": 0.7869,
+      "step": 4282
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017211657327302897,
+      "loss": 0.8121,
+      "step": 4283
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017210412072190542,
+      "loss": 0.7677,
+      "step": 4284
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001720916658414752,
+      "loss": 0.7729,
+      "step": 4285
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017207920863214067,
+      "loss": 0.749,
+      "step": 4286
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017206674909430434,
+      "loss": 0.7378,
+      "step": 4287
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001720542872283686,
+      "loss": 0.7402,
+      "step": 4288
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001720418230347361,
+      "loss": 0.6911,
+      "step": 4289
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017202935651380947,
+      "loss": 0.7487,
+      "step": 4290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017201688766599145,
+      "loss": 0.6953,
+      "step": 4291
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017200441649168485,
+      "loss": 0.6736,
+      "step": 4292
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017199194299129247,
+      "loss": 0.7249,
+      "step": 4293
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017197946716521737,
+      "loss": 0.6796,
+      "step": 4294
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017196698901386252,
+      "loss": 0.7469,
+      "step": 4295
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017195450853763107,
+      "loss": 0.7292,
+      "step": 4296
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017194202573692614,
+      "loss": 0.7906,
+      "step": 4297
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017192954061215102,
+      "loss": 0.7165,
+      "step": 4298
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.000171917053163709,
+      "loss": 0.68,
+      "step": 4299
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017190456339200353,
+      "loss": 0.7707,
+      "step": 4300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001718920712974381,
+      "loss": 0.7054,
+      "step": 4301
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001718795768804162,
+      "loss": 0.7248,
+      "step": 4302
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001718670801413415,
+      "loss": 0.7028,
+      "step": 4303
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017185458108061772,
+      "loss": 0.7084,
+      "step": 4304
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017184207969864863,
+      "loss": 0.7971,
+      "step": 4305
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017182957599583805,
+      "loss": 0.7654,
+      "step": 4306
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017181706997258992,
+      "loss": 0.7064,
+      "step": 4307
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017180456162930828,
+      "loss": 0.7958,
+      "step": 4308
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017179205096639718,
+      "loss": 0.6892,
+      "step": 4309
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001717795379842608,
+      "loss": 0.8037,
+      "step": 4310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001717670226833033,
+      "loss": 0.6956,
+      "step": 4311
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017175450506392908,
+      "loss": 0.7324,
+      "step": 4312
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017174198512654246,
+      "loss": 0.7462,
+      "step": 4313
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001717294628715479,
+      "loss": 0.8464,
+      "step": 4314
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017171693829934993,
+      "loss": 0.7019,
+      "step": 4315
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017170441141035317,
+      "loss": 0.8391,
+      "step": 4316
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017169188220496226,
+      "loss": 0.7351,
+      "step": 4317
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017167935068358202,
+      "loss": 0.7351,
+      "step": 4318
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017166681684661718,
+      "loss": 0.6751,
+      "step": 4319
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017165428069447274,
+      "loss": 0.7195,
+      "step": 4320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017164174222755363,
+      "loss": 0.7506,
+      "step": 4321
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001716292014462649,
+      "loss": 0.6923,
+      "step": 4322
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017161665835101168,
+      "loss": 0.7351,
+      "step": 4323
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001716041129421992,
+      "loss": 0.8036,
+      "step": 4324
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017159156522023267,
+      "loss": 0.7866,
+      "step": 4325
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017157901518551752,
+      "loss": 0.7301,
+      "step": 4326
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017156646283845913,
+      "loss": 0.7746,
+      "step": 4327
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017155390817946305,
+      "loss": 0.8186,
+      "step": 4328
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017154135120893478,
+      "loss": 0.7578,
+      "step": 4329
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017152879192728002,
+      "loss": 0.7295,
+      "step": 4330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017151623033490447,
+      "loss": 0.7535,
+      "step": 4331
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017150366643221393,
+      "loss": 0.7958,
+      "step": 4332
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001714911002196143,
+      "loss": 0.7965,
+      "step": 4333
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001714785316975115,
+      "loss": 0.7743,
+      "step": 4334
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017146596086631158,
+      "loss": 0.6943,
+      "step": 4335
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017145338772642064,
+      "loss": 0.7352,
+      "step": 4336
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001714408122782448,
+      "loss": 0.6326,
+      "step": 4337
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017142823452219038,
+      "loss": 0.7469,
+      "step": 4338
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017141565445866362,
+      "loss": 0.7244,
+      "step": 4339
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.000171403072088071,
+      "loss": 0.7564,
+      "step": 4340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017139048741081892,
+      "loss": 0.8202,
+      "step": 4341
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.000171377900427314,
+      "loss": 0.6964,
+      "step": 4342
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017136531113796276,
+      "loss": 0.7702,
+      "step": 4343
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017135271954317198,
+      "loss": 0.7357,
+      "step": 4344
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001713401256433484,
+      "loss": 0.7455,
+      "step": 4345
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017132752943889887,
+      "loss": 0.7258,
+      "step": 4346
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001713149309302303,
+      "loss": 0.7108,
+      "step": 4347
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017130233011774964,
+      "loss": 0.7307,
+      "step": 4348
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017128972700186404,
+      "loss": 0.7176,
+      "step": 4349
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001712771215829806,
+      "loss": 0.7234,
+      "step": 4350
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001712645138615065,
+      "loss": 0.8093,
+      "step": 4351
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017125190383784908,
+      "loss": 0.7529,
+      "step": 4352
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001712392915124157,
+      "loss": 0.7137,
+      "step": 4353
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017122667688561377,
+      "loss": 0.7474,
+      "step": 4354
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001712140599578508,
+      "loss": 0.7299,
+      "step": 4355
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017120144072953441,
+      "loss": 0.8096,
+      "step": 4356
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017118881920107223,
+      "loss": 0.7291,
+      "step": 4357
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.000171176195372872,
+      "loss": 0.7357,
+      "step": 4358
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001711635692453416,
+      "loss": 0.7234,
+      "step": 4359
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017115094081888882,
+      "loss": 0.7231,
+      "step": 4360
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017113831009392164,
+      "loss": 0.7046,
+      "step": 4361
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001711256770708481,
+      "loss": 0.7473,
+      "step": 4362
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017111304175007632,
+      "loss": 0.6763,
+      "step": 4363
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017110040413201446,
+      "loss": 0.7138,
+      "step": 4364
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017108776421707083,
+      "loss": 0.7548,
+      "step": 4365
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001710751220056537,
+      "loss": 0.7883,
+      "step": 4366
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001710624774981715,
+      "loss": 0.7274,
+      "step": 4367
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001710498306950327,
+      "loss": 0.6949,
+      "step": 4368
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017103718159664578,
+      "loss": 0.7667,
+      "step": 4369
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001710245302034195,
+      "loss": 0.7641,
+      "step": 4370
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017101187651576252,
+      "loss": 0.755,
+      "step": 4371
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017099922053408356,
+      "loss": 0.7505,
+      "step": 4372
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017098656225879148,
+      "loss": 0.7206,
+      "step": 4373
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017097390169029526,
+      "loss": 0.7398,
+      "step": 4374
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017096123882900385,
+      "loss": 0.7117,
+      "step": 4375
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017094857367532632,
+      "loss": 0.6821,
+      "step": 4376
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017093590622967183,
+      "loss": 0.8027,
+      "step": 4377
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001709232364924496,
+      "loss": 0.7603,
+      "step": 4378
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001709105644640689,
+      "loss": 0.6887,
+      "step": 4379
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017089789014493914,
+      "loss": 0.6973,
+      "step": 4380
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001708852135354697,
+      "loss": 0.7662,
+      "step": 4381
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017087253463607017,
+      "loss": 0.7176,
+      "step": 4382
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017085985344715007,
+      "loss": 0.7354,
+      "step": 4383
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017084716996911911,
+      "loss": 0.6403,
+      "step": 4384
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.000170834484202387,
+      "loss": 0.764,
+      "step": 4385
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017082179614736357,
+      "loss": 0.7486,
+      "step": 4386
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017080910580445868,
+      "loss": 0.7429,
+      "step": 4387
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001707964131740823,
+      "loss": 0.737,
+      "step": 4388
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017078371825664446,
+      "loss": 0.8207,
+      "step": 4389
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017077102105255532,
+      "loss": 0.7753,
+      "step": 4390
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017075832156222495,
+      "loss": 0.7339,
+      "step": 4391
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001707456197860637,
+      "loss": 0.7441,
+      "step": 4392
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017073291572448183,
+      "loss": 0.7748,
+      "step": 4393
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017072020937788983,
+      "loss": 0.6995,
+      "step": 4394
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017070750074669804,
+      "loss": 0.7301,
+      "step": 4395
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017069478983131715,
+      "loss": 0.7666,
+      "step": 4396
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001706820766321577,
+      "loss": 0.6438,
+      "step": 4397
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017066936114963044,
+      "loss": 0.7347,
+      "step": 4398
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017065664338414608,
+      "loss": 0.7039,
+      "step": 4399
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001706439233361155,
+      "loss": 0.7016,
+      "step": 4400
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001706312010059496,
+      "loss": 0.7203,
+      "step": 4401
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017061847639405939,
+      "loss": 0.7351,
+      "step": 4402
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017060574950085596,
+      "loss": 0.729,
+      "step": 4403
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001705930203267504,
+      "loss": 0.7687,
+      "step": 4404
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017058028887215394,
+      "loss": 0.7348,
+      "step": 4405
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017056755513747787,
+      "loss": 0.7007,
+      "step": 4406
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017055481912313356,
+      "loss": 0.7054,
+      "step": 4407
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001705420808295324,
+      "loss": 0.8184,
+      "step": 4408
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017052934025708595,
+      "loss": 0.7809,
+      "step": 4409
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017051659740620578,
+      "loss": 0.7911,
+      "step": 4410
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017050385227730352,
+      "loss": 0.7316,
+      "step": 4411
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017049110487079094,
+      "loss": 0.7751,
+      "step": 4412
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017047835518707978,
+      "loss": 0.8124,
+      "step": 4413
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000170465603226582,
+      "loss": 0.6982,
+      "step": 4414
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017045284898970948,
+      "loss": 0.7692,
+      "step": 4415
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017044009247687428,
+      "loss": 0.7316,
+      "step": 4416
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017042733368848848,
+      "loss": 0.6934,
+      "step": 4417
+    },
+    {
+      "epoch": 1.01,
+      "eval_loss": 0.7041641473770142,
+      "eval_runtime": 1184.5539,
+      "eval_samples_per_second": 42.264,
+      "eval_steps_per_second": 10.566,
+      "step": 4417
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017041457262496424,
+      "loss": 0.6994,
+      "step": 4418
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001704018092867138,
+      "loss": 0.7303,
+      "step": 4419
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017038904367414949,
+      "loss": 0.6814,
+      "step": 4420
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001703762757876837,
+      "loss": 0.7598,
+      "step": 4421
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017036350562772893,
+      "loss": 0.7525,
+      "step": 4422
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017035073319469766,
+      "loss": 0.7809,
+      "step": 4423
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001703379584890025,
+      "loss": 0.7656,
+      "step": 4424
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001703251815110562,
+      "loss": 0.7949,
+      "step": 4425
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001703124022612714,
+      "loss": 0.7866,
+      "step": 4426
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017029962074006108,
+      "loss": 0.7684,
+      "step": 4427
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017028683694783804,
+      "loss": 0.7035,
+      "step": 4428
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017027405088501527,
+      "loss": 0.7393,
+      "step": 4429
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017026126255200583,
+      "loss": 0.7206,
+      "step": 4430
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017024847194922285,
+      "loss": 0.7572,
+      "step": 4431
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017023567907707956,
+      "loss": 0.7176,
+      "step": 4432
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017022288393598917,
+      "loss": 0.7166,
+      "step": 4433
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017021008652636508,
+      "loss": 0.7394,
+      "step": 4434
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017019728684862067,
+      "loss": 0.7315,
+      "step": 4435
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001701844849031694,
+      "loss": 0.738,
+      "step": 4436
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017017168069042492,
+      "loss": 0.6766,
+      "step": 4437
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001701588742108008,
+      "loss": 0.6461,
+      "step": 4438
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017014606546471078,
+      "loss": 0.8207,
+      "step": 4439
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001701332544525686,
+      "loss": 0.7553,
+      "step": 4440
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001701204411747882,
+      "loss": 0.6906,
+      "step": 4441
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017010762563178343,
+      "loss": 0.7116,
+      "step": 4442
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001700948078239683,
+      "loss": 0.6754,
+      "step": 4443
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017008198775175695,
+      "loss": 0.7735,
+      "step": 4444
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001700691654155635,
+      "loss": 0.7259,
+      "step": 4445
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017005634081580212,
+      "loss": 0.7872,
+      "step": 4446
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017004351395288718,
+      "loss": 0.7034,
+      "step": 4447
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000170030684827233,
+      "loss": 0.7923,
+      "step": 4448
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017001785343925406,
+      "loss": 0.8353,
+      "step": 4449
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017000501978936482,
+      "loss": 0.6914,
+      "step": 4450
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016999218387797992,
+      "loss": 0.7056,
+      "step": 4451
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016997934570551403,
+      "loss": 0.6592,
+      "step": 4452
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016996650527238181,
+      "loss": 0.7273,
+      "step": 4453
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016995366257899813,
+      "loss": 0.7073,
+      "step": 4454
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016994081762577786,
+      "loss": 0.7629,
+      "step": 4455
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016992797041313597,
+      "loss": 0.7323,
+      "step": 4456
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016991512094148743,
+      "loss": 0.7269,
+      "step": 4457
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016990226921124738,
+      "loss": 0.7291,
+      "step": 4458
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000169889415222831,
+      "loss": 0.7373,
+      "step": 4459
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016987655897665352,
+      "loss": 0.7431,
+      "step": 4460
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001698637004731302,
+      "loss": 0.6702,
+      "step": 4461
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016985083971267654,
+      "loss": 0.7688,
+      "step": 4462
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016983797669570795,
+      "loss": 0.7513,
+      "step": 4463
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016982511142263996,
+      "loss": 0.7799,
+      "step": 4464
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001698122438938882,
+      "loss": 0.7001,
+      "step": 4465
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016979937410986832,
+      "loss": 0.6744,
+      "step": 4466
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001697865020709961,
+      "loss": 0.6786,
+      "step": 4467
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016977362777768738,
+      "loss": 0.7162,
+      "step": 4468
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016976075123035806,
+      "loss": 0.7517,
+      "step": 4469
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001697478724294241,
+      "loss": 0.6957,
+      "step": 4470
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001697349913753015,
+      "loss": 0.6566,
+      "step": 4471
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016972210806840646,
+      "loss": 0.7075,
+      "step": 4472
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016970922250915513,
+      "loss": 0.7304,
+      "step": 4473
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001696963346979638,
+      "loss": 0.7078,
+      "step": 4474
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016968344463524877,
+      "loss": 0.6873,
+      "step": 4475
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016967055232142646,
+      "loss": 0.7399,
+      "step": 4476
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016965765775691337,
+      "loss": 0.7057,
+      "step": 4477
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001696447609421261,
+      "loss": 0.7093,
+      "step": 4478
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016963186187748117,
+      "loss": 0.6943,
+      "step": 4479
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016961896056339536,
+      "loss": 0.7757,
+      "step": 4480
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00016960605700028542,
+      "loss": 0.6866,
+      "step": 4481
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001695931511885682,
+      "loss": 0.658,
+      "step": 4482
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016958024312866064,
+      "loss": 0.7383,
+      "step": 4483
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001695673328209797,
+      "loss": 0.7988,
+      "step": 4484
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016955442026594245,
+      "loss": 0.7045,
+      "step": 4485
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000169541505463966,
+      "loss": 0.722,
+      "step": 4486
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016952858841546764,
+      "loss": 0.7654,
+      "step": 4487
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016951566912086458,
+      "loss": 0.7195,
+      "step": 4488
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016950274758057422,
+      "loss": 0.6172,
+      "step": 4489
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016948982379501394,
+      "loss": 0.7347,
+      "step": 4490
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001694768977646013,
+      "loss": 0.7548,
+      "step": 4491
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016946396948975382,
+      "loss": 0.6978,
+      "step": 4492
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016945103897088917,
+      "loss": 0.7107,
+      "step": 4493
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016943810620842506,
+      "loss": 0.7051,
+      "step": 4494
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001694251712027793,
+      "loss": 0.6452,
+      "step": 4495
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016941223395436968,
+      "loss": 0.7121,
+      "step": 4496
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016939929446361423,
+      "loss": 0.7298,
+      "step": 4497
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001693863527309309,
+      "loss": 0.7501,
+      "step": 4498
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001693734087567378,
+      "loss": 0.7216,
+      "step": 4499
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016936046254145305,
+      "loss": 0.7831,
+      "step": 4500
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001693475140854949,
+      "loss": 0.6855,
+      "step": 4501
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016933456338928164,
+      "loss": 0.7383,
+      "step": 4502
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016932161045323164,
+      "loss": 0.6914,
+      "step": 4503
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016930865527776335,
+      "loss": 0.7126,
+      "step": 4504
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016929569786329525,
+      "loss": 0.7604,
+      "step": 4505
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016928273821024593,
+      "loss": 0.7841,
+      "step": 4506
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016926977631903412,
+      "loss": 0.7275,
+      "step": 4507
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016925681219007846,
+      "loss": 0.7242,
+      "step": 4508
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016924384582379782,
+      "loss": 0.6939,
+      "step": 4509
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016923087722061104,
+      "loss": 0.6783,
+      "step": 4510
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016921790638093706,
+      "loss": 0.6509,
+      "step": 4511
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016920493330519493,
+      "loss": 0.7402,
+      "step": 4512
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001691919579938037,
+      "loss": 0.7496,
+      "step": 4513
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001691789804471826,
+      "loss": 0.7121,
+      "step": 4514
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016916600066575082,
+      "loss": 0.67,
+      "step": 4515
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016915301864992764,
+      "loss": 0.6564,
+      "step": 4516
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001691400344001325,
+      "loss": 0.6752,
+      "step": 4517
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016912704791678483,
+      "loss": 0.7225,
+      "step": 4518
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016911405920030416,
+      "loss": 0.7249,
+      "step": 4519
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001691010682511101,
+      "loss": 0.726,
+      "step": 4520
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001690880750696223,
+      "loss": 0.7546,
+      "step": 4521
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016907507965626049,
+      "loss": 0.7217,
+      "step": 4522
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001690620820114445,
+      "loss": 0.6897,
+      "step": 4523
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001690490821355942,
+      "loss": 0.7549,
+      "step": 4524
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016903608002912955,
+      "loss": 0.6886,
+      "step": 4525
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00016902307569247065,
+      "loss": 0.7055,
+      "step": 4526
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001690100691260375,
+      "loss": 0.7272,
+      "step": 4527
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016899706033025034,
+      "loss": 0.6974,
+      "step": 4528
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001689840493055294,
+      "loss": 0.7435,
+      "step": 4529
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016897103605229496,
+      "loss": 0.6265,
+      "step": 4530
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016895802057096746,
+      "loss": 0.731,
+      "step": 4531
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016894500286196733,
+      "loss": 0.7098,
+      "step": 4532
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016893198292571512,
+      "loss": 0.6377,
+      "step": 4533
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016891896076263148,
+      "loss": 0.6211,
+      "step": 4534
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.000168905936373137,
+      "loss": 0.7304,
+      "step": 4535
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016889290975765245,
+      "loss": 0.7186,
+      "step": 4536
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016887988091659873,
+      "loss": 0.7337,
+      "step": 4537
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016886684985039665,
+      "loss": 0.7052,
+      "step": 4538
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001688538165594672,
+      "loss": 0.825,
+      "step": 4539
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016884078104423145,
+      "loss": 0.7154,
+      "step": 4540
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016882774330511047,
+      "loss": 0.7332,
+      "step": 4541
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016881470334252546,
+      "loss": 0.7093,
+      "step": 4542
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016880166115689762,
+      "loss": 0.7632,
+      "step": 4543
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001687886167486483,
+      "loss": 0.7251,
+      "step": 4544
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.000168775570118199,
+      "loss": 0.7627,
+      "step": 4545
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016876252126597104,
+      "loss": 0.6738,
+      "step": 4546
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016874947019238605,
+      "loss": 0.668,
+      "step": 4547
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001687364168978656,
+      "loss": 0.7307,
+      "step": 4548
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001687233613828314,
+      "loss": 0.7016,
+      "step": 4549
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001687103036477052,
+      "loss": 0.6806,
+      "step": 4550
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016869724369290883,
+      "loss": 0.6595,
+      "step": 4551
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016868418151886415,
+      "loss": 0.7271,
+      "step": 4552
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001686711171259932,
+      "loss": 0.6875,
+      "step": 4553
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016865805051471795,
+      "loss": 0.7032,
+      "step": 4554
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016864498168546054,
+      "loss": 0.6236,
+      "step": 4555
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001686319106386432,
+      "loss": 0.7177,
+      "step": 4556
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001686188373746881,
+      "loss": 0.7782,
+      "step": 4557
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016860576189401766,
+      "loss": 0.6658,
+      "step": 4558
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016859268419705422,
+      "loss": 0.762,
+      "step": 4559
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001685796042842203,
+      "loss": 0.6947,
+      "step": 4560
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016856652215593835,
+      "loss": 0.6746,
+      "step": 4561
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016855343781263112,
+      "loss": 0.6971,
+      "step": 4562
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016854035125472117,
+      "loss": 0.7167,
+      "step": 4563
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016852726248263133,
+      "loss": 0.7997,
+      "step": 4564
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016851417149678444,
+      "loss": 0.7078,
+      "step": 4565
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016850107829760336,
+      "loss": 0.7614,
+      "step": 4566
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016848798288551103,
+      "loss": 0.704,
+      "step": 4567
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001684748852609306,
+      "loss": 0.7294,
+      "step": 4568
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001684617854242851,
+      "loss": 0.7165,
+      "step": 4569
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00016844868337599777,
+      "loss": 0.6535,
+      "step": 4570
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001684355791164918,
+      "loss": 0.7564,
+      "step": 4571
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001684224726461906,
+      "loss": 0.6719,
+      "step": 4572
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001684093639655175,
+      "loss": 0.6908,
+      "step": 4573
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016839625307489607,
+      "loss": 0.6948,
+      "step": 4574
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016838313997474976,
+      "loss": 0.7339,
+      "step": 4575
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001683700246655022,
+      "loss": 0.7143,
+      "step": 4576
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001683569071475771,
+      "loss": 0.6861,
+      "step": 4577
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001683437874213982,
+      "loss": 0.7691,
+      "step": 4578
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016833066548738935,
+      "loss": 0.7382,
+      "step": 4579
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016831754134597447,
+      "loss": 0.7852,
+      "step": 4580
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001683044149975775,
+      "loss": 0.6743,
+      "step": 4581
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016829128644262243,
+      "loss": 0.7501,
+      "step": 4582
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016827815568153348,
+      "loss": 0.6963,
+      "step": 4583
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016826502271473479,
+      "loss": 0.7082,
+      "step": 4584
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001682518875426506,
+      "loss": 0.6999,
+      "step": 4585
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016823875016570527,
+      "loss": 0.7034,
+      "step": 4586
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016822561058432318,
+      "loss": 0.7197,
+      "step": 4587
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016821246879892882,
+      "loss": 0.72,
+      "step": 4588
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016819932480994671,
+      "loss": 0.7049,
+      "step": 4589
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016818617861780148,
+      "loss": 0.7279,
+      "step": 4590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001681730302229178,
+      "loss": 0.788,
+      "step": 4591
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016815987962572043,
+      "loss": 0.7086,
+      "step": 4592
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001681467268266342,
+      "loss": 0.7429,
+      "step": 4593
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.000168133571826084,
+      "loss": 0.6684,
+      "step": 4594
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016812041462449485,
+      "loss": 0.6941,
+      "step": 4595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016810725522229173,
+      "loss": 0.6956,
+      "step": 4596
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001680940936198997,
+      "loss": 0.7047,
+      "step": 4597
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001680809298177441,
+      "loss": 0.7021,
+      "step": 4598
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016806776381625004,
+      "loss": 0.7037,
+      "step": 4599
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016805459561584294,
+      "loss": 0.7009,
+      "step": 4600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016804142521694812,
+      "loss": 0.706,
+      "step": 4601
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016802825261999109,
+      "loss": 0.6287,
+      "step": 4602
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001680150778253974,
+      "loss": 0.6344,
+      "step": 4603
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016800190083359262,
+      "loss": 0.7687,
+      "step": 4604
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016798872164500245,
+      "loss": 0.7098,
+      "step": 4605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016797554026005265,
+      "loss": 0.7146,
+      "step": 4606
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016796235667916902,
+      "loss": 0.724,
+      "step": 4607
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016794917090277747,
+      "loss": 0.6858,
+      "step": 4608
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016793598293130395,
+      "loss": 0.7047,
+      "step": 4609
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016792279276517448,
+      "loss": 0.6871,
+      "step": 4610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016790960040481522,
+      "loss": 0.6471,
+      "step": 4611
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001678964058506523,
+      "loss": 0.6673,
+      "step": 4612
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00016788320910311196,
+      "loss": 0.7347,
+      "step": 4613
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016787001016262053,
+      "loss": 0.7735,
+      "step": 4614
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016785680902960443,
+      "loss": 0.7158,
+      "step": 4615
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001678436057044901,
+      "loss": 0.7207,
+      "step": 4616
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016783040018770407,
+      "loss": 0.7078,
+      "step": 4617
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001678171924796729,
+      "loss": 0.7421,
+      "step": 4618
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001678039825808233,
+      "loss": 0.7153,
+      "step": 4619
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016779077049158204,
+      "loss": 0.7658,
+      "step": 4620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016777755621237588,
+      "loss": 0.7445,
+      "step": 4621
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016776433974363173,
+      "loss": 0.705,
+      "step": 4622
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016775112108577655,
+      "loss": 0.6745,
+      "step": 4623
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016773790023923737,
+      "loss": 0.7111,
+      "step": 4624
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016772467720444121,
+      "loss": 0.6843,
+      "step": 4625
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016771145198181539,
+      "loss": 0.8,
+      "step": 4626
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016769822457178704,
+      "loss": 0.6948,
+      "step": 4627
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016768499497478345,
+      "loss": 0.7483,
+      "step": 4628
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016767176319123207,
+      "loss": 0.7461,
+      "step": 4629
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001676585292215603,
+      "loss": 0.7626,
+      "step": 4630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016764529306619567,
+      "loss": 0.7063,
+      "step": 4631
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001676320547255658,
+      "loss": 0.693,
+      "step": 4632
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001676188142000983,
+      "loss": 0.6765,
+      "step": 4633
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016760557149022096,
+      "loss": 0.6984,
+      "step": 4634
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016759232659636152,
+      "loss": 0.7021,
+      "step": 4635
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016757907951894793,
+      "loss": 0.7517,
+      "step": 4636
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016756583025840803,
+      "loss": 0.7451,
+      "step": 4637
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016755257881516996,
+      "loss": 0.6922,
+      "step": 4638
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001675393251896617,
+      "loss": 0.6471,
+      "step": 4639
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016752606938231148,
+      "loss": 0.7302,
+      "step": 4640
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016751281139354743,
+      "loss": 0.7126,
+      "step": 4641
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016749955122379796,
+      "loss": 0.7872,
+      "step": 4642
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016748628887349137,
+      "loss": 0.6827,
+      "step": 4643
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016747302434305615,
+      "loss": 0.6267,
+      "step": 4644
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001674597576329207,
+      "loss": 0.6523,
+      "step": 4645
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001674464887435137,
+      "loss": 0.6794,
+      "step": 4646
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016743321767526375,
+      "loss": 0.7334,
+      "step": 4647
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001674199444285996,
+      "loss": 0.7147,
+      "step": 4648
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016740666900395002,
+      "loss": 0.665,
+      "step": 4649
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016739339140174386,
+      "loss": 0.8183,
+      "step": 4650
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016738011162241009,
+      "loss": 0.7663,
+      "step": 4651
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016736682966637766,
+      "loss": 0.6982,
+      "step": 4652
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016735354553407562,
+      "loss": 0.719,
+      "step": 4653
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001673402592259332,
+      "loss": 0.7122,
+      "step": 4654
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016732697074237955,
+      "loss": 0.7002,
+      "step": 4655
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00016731368008384397,
+      "loss": 0.701,
+      "step": 4656
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001673003872507558,
+      "loss": 0.7332,
+      "step": 4657
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001672870922435445,
+      "loss": 0.7698,
+      "step": 4658
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016727379506263949,
+      "loss": 0.7181,
+      "step": 4659
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001672604957084704,
+      "loss": 0.7333,
+      "step": 4660
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016724719418146684,
+      "loss": 0.6947,
+      "step": 4661
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016723389048205848,
+      "loss": 0.7313,
+      "step": 4662
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016722058461067512,
+      "loss": 0.7868,
+      "step": 4663
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016720727656774665,
+      "loss": 0.6462,
+      "step": 4664
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001671939663537029,
+      "loss": 0.7059,
+      "step": 4665
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001671806539689739,
+      "loss": 0.6919,
+      "step": 4666
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001671673394139897,
+      "loss": 0.6784,
+      "step": 4667
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001671540226891804,
+      "loss": 0.7092,
+      "step": 4668
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016714070379497624,
+      "loss": 0.7622,
+      "step": 4669
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016712738273180744,
+      "loss": 0.6901,
+      "step": 4670
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016711405950010433,
+      "loss": 0.7197,
+      "step": 4671
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016710073410029734,
+      "loss": 0.6817,
+      "step": 4672
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016708740653281694,
+      "loss": 0.8515,
+      "step": 4673
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016707407679809368,
+      "loss": 0.6889,
+      "step": 4674
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016706074489655816,
+      "loss": 0.6871,
+      "step": 4675
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016704741082864107,
+      "loss": 0.715,
+      "step": 4676
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016703407459477316,
+      "loss": 0.6691,
+      "step": 4677
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016702073619538526,
+      "loss": 0.729,
+      "step": 4678
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016700739563090826,
+      "loss": 0.734,
+      "step": 4679
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001669940529017731,
+      "loss": 0.7142,
+      "step": 4680
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016698070800841088,
+      "loss": 0.6852,
+      "step": 4681
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016696736095125263,
+      "loss": 0.7037,
+      "step": 4682
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001669540117307296,
+      "loss": 0.7073,
+      "step": 4683
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016694066034727293,
+      "loss": 0.7498,
+      "step": 4684
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016692730680131402,
+      "loss": 0.6475,
+      "step": 4685
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001669139510932842,
+      "loss": 0.7141,
+      "step": 4686
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.000166900593223615,
+      "loss": 0.7056,
+      "step": 4687
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016688723319273786,
+      "loss": 0.7329,
+      "step": 4688
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016687387100108438,
+      "loss": 0.7781,
+      "step": 4689
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016686050664908625,
+      "loss": 0.6342,
+      "step": 4690
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016684714013717522,
+      "loss": 0.6589,
+      "step": 4691
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001668337714657831,
+      "loss": 0.7636,
+      "step": 4692
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016682040063534167,
+      "loss": 0.7081,
+      "step": 4693
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016680702764628299,
+      "loss": 0.6856,
+      "step": 4694
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016679365249903894,
+      "loss": 0.7672,
+      "step": 4695
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016678027519404176,
+      "loss": 0.6615,
+      "step": 4696
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016676689573172346,
+      "loss": 0.6022,
+      "step": 4697
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016675351411251635,
+      "loss": 0.6697,
+      "step": 4698
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016674013033685266,
+      "loss": 0.6978,
+      "step": 4699
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016672674440516478,
+      "loss": 0.7176,
+      "step": 4700
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00016671335631788514,
+      "loss": 0.6828,
+      "step": 4701
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016669996607544625,
+      "loss": 0.7002,
+      "step": 4702
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016668657367828064,
+      "loss": 0.6938,
+      "step": 4703
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016667317912682098,
+      "loss": 0.7344,
+      "step": 4704
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016665978242149998,
+      "loss": 0.7557,
+      "step": 4705
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016664638356275039,
+      "loss": 0.7141,
+      "step": 4706
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016663298255100508,
+      "loss": 0.6875,
+      "step": 4707
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016661957938669697,
+      "loss": 0.7998,
+      "step": 4708
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.000166606174070259,
+      "loss": 0.6424,
+      "step": 4709
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001665927666021243,
+      "loss": 0.6972,
+      "step": 4710
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016657935698272596,
+      "loss": 0.6892,
+      "step": 4711
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016656594521249714,
+      "loss": 0.724,
+      "step": 4712
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001665525312918712,
+      "loss": 0.7072,
+      "step": 4713
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016653911522128133,
+      "loss": 0.7288,
+      "step": 4714
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016652569700116105,
+      "loss": 0.7332,
+      "step": 4715
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016651227663194379,
+      "loss": 0.7066,
+      "step": 4716
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016649885411406312,
+      "loss": 0.6996,
+      "step": 4717
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001664854294479526,
+      "loss": 0.6813,
+      "step": 4718
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016647200263404595,
+      "loss": 0.7396,
+      "step": 4719
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001664585736727769,
+      "loss": 0.7091,
+      "step": 4720
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001664451425645793,
+      "loss": 0.678,
+      "step": 4721
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016643170930988698,
+      "loss": 0.752,
+      "step": 4722
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016641827390913397,
+      "loss": 0.6679,
+      "step": 4723
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016640483636275421,
+      "loss": 0.7198,
+      "step": 4724
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001663913966711819,
+      "loss": 0.7204,
+      "step": 4725
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016637795483485113,
+      "loss": 0.6636,
+      "step": 4726
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016636451085419617,
+      "loss": 0.734,
+      "step": 4727
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016635106472965132,
+      "loss": 0.6984,
+      "step": 4728
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016633761646165095,
+      "loss": 0.7356,
+      "step": 4729
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001663241660506295,
+      "loss": 0.6628,
+      "step": 4730
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016631071349702147,
+      "loss": 0.6456,
+      "step": 4731
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001662972588012615,
+      "loss": 0.6591,
+      "step": 4732
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016628380196378415,
+      "loss": 0.7858,
+      "step": 4733
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001662703429850242,
+      "loss": 0.7137,
+      "step": 4734
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016625688186541643,
+      "loss": 0.6909,
+      "step": 4735
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001662434186053957,
+      "loss": 0.684,
+      "step": 4736
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001662299532053969,
+      "loss": 0.6807,
+      "step": 4737
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016621648566585512,
+      "loss": 0.7723,
+      "step": 4738
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016620301598720531,
+      "loss": 0.7128,
+      "step": 4739
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016618954416988268,
+      "loss": 0.7044,
+      "step": 4740
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016617607021432242,
+      "loss": 0.6697,
+      "step": 4741
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016616259412095976,
+      "loss": 0.7171,
+      "step": 4742
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001661491158902301,
+      "loss": 0.7634,
+      "step": 4743
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00016613563552256882,
+      "loss": 0.7093,
+      "step": 4744
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016612215301841144,
+      "loss": 0.6725,
+      "step": 4745
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016610866837819347,
+      "loss": 0.7345,
+      "step": 4746
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001660951816023505,
+      "loss": 0.7005,
+      "step": 4747
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016608169269131828,
+      "loss": 0.764,
+      "step": 4748
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016606820164553256,
+      "loss": 0.7221,
+      "step": 4749
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001660547084654291,
+      "loss": 0.7167,
+      "step": 4750
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001660412131514439,
+      "loss": 0.7315,
+      "step": 4751
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016602771570401278,
+      "loss": 0.7206,
+      "step": 4752
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016601421612357192,
+      "loss": 0.7276,
+      "step": 4753
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016600071441055732,
+      "loss": 0.7989,
+      "step": 4754
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016598721056540518,
+      "loss": 0.7231,
+      "step": 4755
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016597370458855176,
+      "loss": 0.7819,
+      "step": 4756
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016596019648043337,
+      "loss": 0.7889,
+      "step": 4757
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016594668624148634,
+      "loss": 0.7273,
+      "step": 4758
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016593317387214715,
+      "loss": 0.6926,
+      "step": 4759
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016591965937285227,
+      "loss": 0.7442,
+      "step": 4760
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016590614274403836,
+      "loss": 0.6939,
+      "step": 4761
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.000165892623986142,
+      "loss": 0.6874,
+      "step": 4762
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016587910309959995,
+      "loss": 0.6719,
+      "step": 4763
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016586558008484898,
+      "loss": 0.6933,
+      "step": 4764
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.000165852054942326,
+      "loss": 0.7505,
+      "step": 4765
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016583852767246783,
+      "loss": 0.7119,
+      "step": 4766
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001658249982757116,
+      "loss": 0.6781,
+      "step": 4767
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016581146675249424,
+      "loss": 0.7399,
+      "step": 4768
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016579793310325295,
+      "loss": 0.7521,
+      "step": 4769
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016578439732842495,
+      "loss": 0.7989,
+      "step": 4770
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016577085942844747,
+      "loss": 0.7144,
+      "step": 4771
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016575731940375788,
+      "loss": 0.6519,
+      "step": 4772
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016574377725479358,
+      "loss": 0.7534,
+      "step": 4773
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016573023298199202,
+      "loss": 0.6895,
+      "step": 4774
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016571668658579077,
+      "loss": 0.6762,
+      "step": 4775
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016570313806662742,
+      "loss": 0.7228,
+      "step": 4776
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016568958742493967,
+      "loss": 0.7016,
+      "step": 4777
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016567603466116529,
+      "loss": 0.7039,
+      "step": 4778
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016566247977574204,
+      "loss": 0.6725,
+      "step": 4779
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016564892276910788,
+      "loss": 0.7375,
+      "step": 4780
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016563536364170073,
+      "loss": 0.6924,
+      "step": 4781
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001656218023939586,
+      "loss": 0.7037,
+      "step": 4782
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001656082390263196,
+      "loss": 0.7303,
+      "step": 4783
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001655946735392219,
+      "loss": 0.7016,
+      "step": 4784
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001655811059331037,
+      "loss": 0.6994,
+      "step": 4785
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016556753620840332,
+      "loss": 0.7314,
+      "step": 4786
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016555396436555912,
+      "loss": 0.7515,
+      "step": 4787
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00016554039040500955,
+      "loss": 0.7122,
+      "step": 4788
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001655268143271931,
+      "loss": 0.7004,
+      "step": 4789
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016551323613254833,
+      "loss": 0.6996,
+      "step": 4790
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001654996558215139,
+      "loss": 0.7714,
+      "step": 4791
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016548607339452853,
+      "loss": 0.7894,
+      "step": 4792
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016547248885203096,
+      "loss": 0.6979,
+      "step": 4793
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016545890219446006,
+      "loss": 0.768,
+      "step": 4794
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016544531342225474,
+      "loss": 0.7714,
+      "step": 4795
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016543172253585396,
+      "loss": 0.7589,
+      "step": 4796
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016541812953569686,
+      "loss": 0.6673,
+      "step": 4797
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016540453442222245,
+      "loss": 0.7562,
+      "step": 4798
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016539093719586994,
+      "loss": 0.7669,
+      "step": 4799
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016537733785707866,
+      "loss": 0.6993,
+      "step": 4800
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001653637364062878,
+      "loss": 0.7813,
+      "step": 4801
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016535013284393687,
+      "loss": 0.7702,
+      "step": 4802
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001653365271704653,
+      "loss": 0.6699,
+      "step": 4803
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016532291938631258,
+      "loss": 0.6634,
+      "step": 4804
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016530930949191835,
+      "loss": 0.6621,
+      "step": 4805
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016529569748772224,
+      "loss": 0.73,
+      "step": 4806
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.000165282083374164,
+      "loss": 0.71,
+      "step": 4807
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016526846715168346,
+      "loss": 0.6652,
+      "step": 4808
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016525484882072043,
+      "loss": 0.7153,
+      "step": 4809
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001652412283817149,
+      "loss": 0.7208,
+      "step": 4810
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016522760583510682,
+      "loss": 0.7331,
+      "step": 4811
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016521398118133628,
+      "loss": 0.7141,
+      "step": 4812
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001652003544208435,
+      "loss": 0.6897,
+      "step": 4813
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016518672555406857,
+      "loss": 0.6774,
+      "step": 4814
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016517309458145184,
+      "loss": 0.6695,
+      "step": 4815
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016515946150343365,
+      "loss": 0.7138,
+      "step": 4816
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001651458263204544,
+      "loss": 0.6195,
+      "step": 4817
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016513218903295456,
+      "loss": 0.6753,
+      "step": 4818
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001651185496413747,
+      "loss": 0.7009,
+      "step": 4819
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001651049081461554,
+      "loss": 0.7544,
+      "step": 4820
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001650912645477374,
+      "loss": 0.7383,
+      "step": 4821
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016507761884656144,
+      "loss": 0.7099,
+      "step": 4822
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001650639710430683,
+      "loss": 0.7044,
+      "step": 4823
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001650503211376989,
+      "loss": 0.6749,
+      "step": 4824
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016503666913089422,
+      "loss": 0.737,
+      "step": 4825
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016502301502309522,
+      "loss": 0.7983,
+      "step": 4826
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001650093588147431,
+      "loss": 0.7251,
+      "step": 4827
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016499570050627888,
+      "loss": 0.7462,
+      "step": 4828
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001649820400981439,
+      "loss": 0.7405,
+      "step": 4829
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001649683775907794,
+      "loss": 0.7187,
+      "step": 4830
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016495471298462678,
+      "loss": 0.7184,
+      "step": 4831
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00016494104628012743,
+      "loss": 0.7356,
+      "step": 4832
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00016492737747772287,
+      "loss": 0.7318,
+      "step": 4833
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001649137065778547,
+      "loss": 0.7213,
+      "step": 4834
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001649000335809645,
+      "loss": 0.6772,
+      "step": 4835
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.000164886358487494,
+      "loss": 0.6934,
+      "step": 4836
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00016487268129788501,
+      "loss": 0.7773,
+      "step": 4837
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001648590020125793,
+      "loss": 0.691,
+      "step": 4838
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001648453206320188,
+      "loss": 0.7202,
+      "step": 4839
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001648316371566455,
+      "loss": 0.741,
+      "step": 4840
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00016481795158690142,
+      "loss": 0.656,
+      "step": 4841
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00016480426392322865,
+      "loss": 0.7017,
+      "step": 4842
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00016479057416606945,
+      "loss": 0.6782,
+      "step": 4843
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00016477688231586595,
+      "loss": 0.6878,
+      "step": 4844
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016396470087619462,
+      "loss": 0.6767,
+      "step": 4845
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016395071774281524,
+      "loss": 0.7063,
+      "step": 4846
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001639367324934625,
+      "loss": 0.689,
+      "step": 4847
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016392274512859914,
+      "loss": 0.6851,
+      "step": 4848
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016390875564868787,
+      "loss": 0.7232,
+      "step": 4849
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016389476405419173,
+      "loss": 0.6756,
+      "step": 4850
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016388077034557355,
+      "loss": 0.7767,
+      "step": 4851
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001638667745232964,
+      "loss": 0.7715,
+      "step": 4852
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016385277658782332,
+      "loss": 0.6857,
+      "step": 4853
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016383877653961752,
+      "loss": 0.6884,
+      "step": 4854
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001638247743791422,
+      "loss": 0.6448,
+      "step": 4855
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016381077010686068,
+      "loss": 0.6969,
+      "step": 4856
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00016379676372323628,
+      "loss": 0.7312,
+      "step": 4857
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001637827552287325,
+      "loss": 0.7356,
+      "step": 4858
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016376874462381279,
+      "loss": 0.6286,
+      "step": 4859
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016375473190894074,
+      "loss": 0.6425,
+      "step": 4860
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016374071708458003,
+      "loss": 0.7228,
+      "step": 4861
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016372670015119434,
+      "loss": 0.7349,
+      "step": 4862
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001637126811092475,
+      "loss": 0.7657,
+      "step": 4863
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016369865995920332,
+      "loss": 0.7099,
+      "step": 4864
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016368463670152572,
+      "loss": 0.6672,
+      "step": 4865
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016367061133667875,
+      "loss": 0.7146,
+      "step": 4866
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016365658386512642,
+      "loss": 0.6883,
+      "step": 4867
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001636425542873329,
+      "loss": 0.7156,
+      "step": 4868
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016362852260376237,
+      "loss": 0.6752,
+      "step": 4869
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016361448881487914,
+      "loss": 0.7177,
+      "step": 4870
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001636004529211475,
+      "loss": 0.6808,
+      "step": 4871
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001635864149230319,
+      "loss": 0.6523,
+      "step": 4872
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016357237482099684,
+      "loss": 0.67,
+      "step": 4873
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001635583326155068,
+      "loss": 0.7223,
+      "step": 4874
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016354428830702648,
+      "loss": 0.7343,
+      "step": 4875
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016353024189602055,
+      "loss": 0.634,
+      "step": 4876
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016351619338295375,
+      "loss": 0.7267,
+      "step": 4877
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016350214276829094,
+      "loss": 0.6076,
+      "step": 4878
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.000163488090052497,
+      "loss": 0.7105,
+      "step": 4879
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001634740352360369,
+      "loss": 0.6441,
+      "step": 4880
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001634599783193757,
+      "loss": 0.6678,
+      "step": 4881
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016344591930297849,
+      "loss": 0.6645,
+      "step": 4882
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016343185818731046,
+      "loss": 0.7744,
+      "step": 4883
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016341779497283684,
+      "loss": 0.7541,
+      "step": 4884
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016340372966002295,
+      "loss": 0.7607,
+      "step": 4885
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016338966224933423,
+      "loss": 0.6787,
+      "step": 4886
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016337559274123606,
+      "loss": 0.6753,
+      "step": 4887
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.000163361521136194,
+      "loss": 0.7324,
+      "step": 4888
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016334744743467364,
+      "loss": 0.7214,
+      "step": 4889
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016333337163714067,
+      "loss": 0.7296,
+      "step": 4890
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016331929374406077,
+      "loss": 0.7356,
+      "step": 4891
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001633052137558998,
+      "loss": 0.7163,
+      "step": 4892
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001632911316731236,
+      "loss": 0.7057,
+      "step": 4893
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001632770474961981,
+      "loss": 0.7125,
+      "step": 4894
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016326296122558937,
+      "loss": 0.7692,
+      "step": 4895
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016324887286176342,
+      "loss": 0.7276,
+      "step": 4896
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016323478240518643,
+      "loss": 0.7724,
+      "step": 4897
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016322068985632462,
+      "loss": 0.7055,
+      "step": 4898
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016320659521564432,
+      "loss": 0.6612,
+      "step": 4899
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016319249848361183,
+      "loss": 0.6913,
+      "step": 4900
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016317839966069355,
+      "loss": 0.6931,
+      "step": 4901
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00016316429874735605,
+      "loss": 0.7262,
+      "step": 4902
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016315019574406585,
+      "loss": 0.7034,
+      "step": 4903
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001631360906512896,
+      "loss": 0.6597,
+      "step": 4904
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016312198346949398,
+      "loss": 0.7341,
+      "step": 4905
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001631078741991458,
+      "loss": 0.7197,
+      "step": 4906
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016309376284071188,
+      "loss": 0.7013,
+      "step": 4907
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016307964939465914,
+      "loss": 0.7509,
+      "step": 4908
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016306553386145453,
+      "loss": 0.7007,
+      "step": 4909
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016305141624156514,
+      "loss": 0.709,
+      "step": 4910
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016303729653545807,
+      "loss": 0.6722,
+      "step": 4911
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001630231747436005,
+      "loss": 0.7089,
+      "step": 4912
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001630090508664597,
+      "loss": 0.6099,
+      "step": 4913
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016299492490450297,
+      "loss": 0.656,
+      "step": 4914
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016298079685819774,
+      "loss": 0.6955,
+      "step": 4915
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001629666667280114,
+      "loss": 0.7347,
+      "step": 4916
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001629525345144116,
+      "loss": 0.6868,
+      "step": 4917
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016293840021786585,
+      "loss": 0.7249,
+      "step": 4918
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016292426383884183,
+      "loss": 0.7746,
+      "step": 4919
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001629101253778073,
+      "loss": 0.6645,
+      "step": 4920
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016289598483523008,
+      "loss": 0.6745,
+      "step": 4921
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.000162881842211578,
+      "loss": 0.6812,
+      "step": 4922
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016286769750731902,
+      "loss": 0.7217,
+      "step": 4923
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001628535507229212,
+      "loss": 0.6865,
+      "step": 4924
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016283940185885255,
+      "loss": 0.7031,
+      "step": 4925
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001628252509155813,
+      "loss": 0.6997,
+      "step": 4926
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001628110978935756,
+      "loss": 0.7621,
+      "step": 4927
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016279694279330376,
+      "loss": 0.7119,
+      "step": 4928
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016278278561523415,
+      "loss": 0.7202,
+      "step": 4929
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016276862635983518,
+      "loss": 0.753,
+      "step": 4930
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001627544650275754,
+      "loss": 0.6849,
+      "step": 4931
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016274030161892328,
+      "loss": 0.7155,
+      "step": 4932
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016272613613434756,
+      "loss": 0.6993,
+      "step": 4933
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016271196857431684,
+      "loss": 0.7003,
+      "step": 4934
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016269779893929992,
+      "loss": 0.7397,
+      "step": 4935
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016268362722976569,
+      "loss": 0.685,
+      "step": 4936
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.000162669453446183,
+      "loss": 0.7012,
+      "step": 4937
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016265527758902088,
+      "loss": 0.7341,
+      "step": 4938
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001626410996587483,
+      "loss": 0.727,
+      "step": 4939
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016262691965583445,
+      "loss": 0.6624,
+      "step": 4940
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016261273758074845,
+      "loss": 0.7152,
+      "step": 4941
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016259855343395957,
+      "loss": 0.7383,
+      "step": 4942
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016258436721593714,
+      "loss": 0.6973,
+      "step": 4943
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016257017892715059,
+      "loss": 0.6731,
+      "step": 4944
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00016255598856806927,
+      "loss": 0.7559,
+      "step": 4945
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016254179613916278,
+      "loss": 0.7351,
+      "step": 4946
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001625276016409007,
+      "loss": 0.7269,
+      "step": 4947
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001625134050737527,
+      "loss": 0.6718,
+      "step": 4948
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001624992064381885,
+      "loss": 0.7901,
+      "step": 4949
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016248500573467784,
+      "loss": 0.7393,
+      "step": 4950
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001624708029636907,
+      "loss": 0.7094,
+      "step": 4951
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016245659812569692,
+      "loss": 0.764,
+      "step": 4952
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016244239122116655,
+      "loss": 0.6714,
+      "step": 4953
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001624281822505697,
+      "loss": 0.7573,
+      "step": 4954
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001624139712143764,
+      "loss": 0.7032,
+      "step": 4955
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016239975811305692,
+      "loss": 0.6907,
+      "step": 4956
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016238554294708158,
+      "loss": 0.6691,
+      "step": 4957
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001623713257169207,
+      "loss": 0.7128,
+      "step": 4958
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001623571064230446,
+      "loss": 0.6998,
+      "step": 4959
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016234288506592387,
+      "loss": 0.6842,
+      "step": 4960
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016232866164602906,
+      "loss": 0.7049,
+      "step": 4961
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016231443616383074,
+      "loss": 0.7541,
+      "step": 4962
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001623002086197996,
+      "loss": 0.659,
+      "step": 4963
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001622859790144064,
+      "loss": 0.7129,
+      "step": 4964
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.000162271747348122,
+      "loss": 0.7052,
+      "step": 4965
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016225751362141722,
+      "loss": 0.6795,
+      "step": 4966
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016224327783476313,
+      "loss": 0.6632,
+      "step": 4967
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016222903998863063,
+      "loss": 0.7449,
+      "step": 4968
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016221480008349088,
+      "loss": 0.6746,
+      "step": 4969
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016220055811981508,
+      "loss": 0.6817,
+      "step": 4970
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016218631409807438,
+      "loss": 0.7151,
+      "step": 4971
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016217206801874011,
+      "loss": 0.7452,
+      "step": 4972
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001621578198822837,
+      "loss": 0.7144,
+      "step": 4973
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016214356968917648,
+      "loss": 0.6005,
+      "step": 4974
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016212931743989003,
+      "loss": 0.7699,
+      "step": 4975
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016211506313489591,
+      "loss": 0.7623,
+      "step": 4976
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016210080677466575,
+      "loss": 0.744,
+      "step": 4977
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016208654835967124,
+      "loss": 0.7207,
+      "step": 4978
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001620722878903842,
+      "loss": 0.7138,
+      "step": 4979
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016205802536727642,
+      "loss": 0.7454,
+      "step": 4980
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016204376079081986,
+      "loss": 0.674,
+      "step": 4981
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016202949416148647,
+      "loss": 0.7283,
+      "step": 4982
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016201522547974832,
+      "loss": 0.6529,
+      "step": 4983
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016200095474607753,
+      "loss": 0.7559,
+      "step": 4984
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016198668196094627,
+      "loss": 0.7466,
+      "step": 4985
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016197240712482674,
+      "loss": 0.7058,
+      "step": 4986
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016195813023819135,
+      "loss": 0.7211,
+      "step": 4987
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00016194385130151244,
+      "loss": 0.6863,
+      "step": 4988
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001619295703152625,
+      "loss": 0.7081,
+      "step": 4989
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000161915287279914,
+      "loss": 0.6963,
+      "step": 4990
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016190100219593955,
+      "loss": 0.7649,
+      "step": 4991
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016188671506381182,
+      "loss": 0.6973,
+      "step": 4992
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016187242588400354,
+      "loss": 0.6272,
+      "step": 4993
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016185813465698749,
+      "loss": 0.7467,
+      "step": 4994
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016184384138323655,
+      "loss": 0.6663,
+      "step": 4995
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016182954606322364,
+      "loss": 0.6643,
+      "step": 4996
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016181524869742173,
+      "loss": 0.7302,
+      "step": 4997
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001618009492863039,
+      "loss": 0.7521,
+      "step": 4998
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016178664783034332,
+      "loss": 0.6421,
+      "step": 4999
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016177234433001315,
+      "loss": 0.7345,
+      "step": 5000
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016175803878578667,
+      "loss": 0.7065,
+      "step": 5001
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016174373119813722,
+      "loss": 0.7296,
+      "step": 5002
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001617294215675382,
+      "loss": 0.6596,
+      "step": 5003
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001617151098944631,
+      "loss": 0.7564,
+      "step": 5004
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016170079617938541,
+      "loss": 0.6629,
+      "step": 5005
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016168648042277879,
+      "loss": 0.67,
+      "step": 5006
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016167216262511687,
+      "loss": 0.6711,
+      "step": 5007
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001616578427868734,
+      "loss": 0.747,
+      "step": 5008
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001616435209085222,
+      "loss": 0.7235,
+      "step": 5009
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016162919699053716,
+      "loss": 0.636,
+      "step": 5010
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016161487103339222,
+      "loss": 0.7371,
+      "step": 5011
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016160054303756137,
+      "loss": 0.7285,
+      "step": 5012
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016158621300351865,
+      "loss": 0.6999,
+      "step": 5013
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001615718809317383,
+      "loss": 0.7371,
+      "step": 5014
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016155754682269446,
+      "loss": 0.704,
+      "step": 5015
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016154321067686148,
+      "loss": 0.716,
+      "step": 5016
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016152887249471363,
+      "loss": 0.7309,
+      "step": 5017
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016151453227672538,
+      "loss": 0.7569,
+      "step": 5018
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016150019002337117,
+      "loss": 0.7775,
+      "step": 5019
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016148584573512557,
+      "loss": 0.6367,
+      "step": 5020
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016147149941246323,
+      "loss": 0.6723,
+      "step": 5021
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001614571510558588,
+      "loss": 0.7602,
+      "step": 5022
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000161442800665787,
+      "loss": 0.73,
+      "step": 5023
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016142844824272272,
+      "loss": 0.6453,
+      "step": 5024
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001614140937871408,
+      "loss": 0.7176,
+      "step": 5025
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016139973729951623,
+      "loss": 0.7047,
+      "step": 5026
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000161385378780324,
+      "loss": 0.7218,
+      "step": 5027
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001613710182300392,
+      "loss": 0.7208,
+      "step": 5028
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000161356655649137,
+      "loss": 0.7123,
+      "step": 5029
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001613422910380926,
+      "loss": 0.7845,
+      "step": 5030
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00016132792439738132,
+      "loss": 0.7122,
+      "step": 5031
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016131355572747853,
+      "loss": 0.6986,
+      "step": 5032
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016129918502885956,
+      "loss": 0.7365,
+      "step": 5033
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000161284812302,
+      "loss": 0.7768,
+      "step": 5034
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001612704375473754,
+      "loss": 0.6867,
+      "step": 5035
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016125606076546135,
+      "loss": 0.6514,
+      "step": 5036
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016124168195673355,
+      "loss": 0.7027,
+      "step": 5037
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016122730112166778,
+      "loss": 0.7177,
+      "step": 5038
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016121291826073984,
+      "loss": 0.7192,
+      "step": 5039
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001611985333744256,
+      "loss": 0.6569,
+      "step": 5040
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001611841464632011,
+      "loss": 0.7497,
+      "step": 5041
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016116975752754234,
+      "loss": 0.7354,
+      "step": 5042
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016115536656792539,
+      "loss": 0.6701,
+      "step": 5043
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016114097358482638,
+      "loss": 0.7228,
+      "step": 5044
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016112657857872162,
+      "loss": 0.7348,
+      "step": 5045
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016111218155008734,
+      "loss": 0.6682,
+      "step": 5046
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016109778249939996,
+      "loss": 0.7288,
+      "step": 5047
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016108338142713585,
+      "loss": 0.7426,
+      "step": 5048
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016106897833377155,
+      "loss": 0.6927,
+      "step": 5049
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001610545732197836,
+      "loss": 0.7271,
+      "step": 5050
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016104016608564861,
+      "loss": 0.6573,
+      "step": 5051
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016102575693184332,
+      "loss": 0.7282,
+      "step": 5052
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001610113457588445,
+      "loss": 0.7459,
+      "step": 5053
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016099693256712896,
+      "loss": 0.6926,
+      "step": 5054
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016098251735717357,
+      "loss": 0.7193,
+      "step": 5055
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001609681001294553,
+      "loss": 0.7293,
+      "step": 5056
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016095368088445125,
+      "loss": 0.6974,
+      "step": 5057
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016093925962263846,
+      "loss": 0.727,
+      "step": 5058
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001609248363444941,
+      "loss": 0.7567,
+      "step": 5059
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001609104110504954,
+      "loss": 0.7157,
+      "step": 5060
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016089598374111966,
+      "loss": 0.6979,
+      "step": 5061
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016088155441684423,
+      "loss": 0.7012,
+      "step": 5062
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016086712307814658,
+      "loss": 0.7571,
+      "step": 5063
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016085268972550418,
+      "loss": 0.6612,
+      "step": 5064
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001608382543593946,
+      "loss": 0.6884,
+      "step": 5065
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016082381698029547,
+      "loss": 0.6944,
+      "step": 5066
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016080937758868448,
+      "loss": 0.7226,
+      "step": 5067
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016079493618503942,
+      "loss": 0.7308,
+      "step": 5068
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001607804927698381,
+      "loss": 0.7843,
+      "step": 5069
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016076604734355842,
+      "loss": 0.7666,
+      "step": 5070
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016075159990667834,
+      "loss": 0.6716,
+      "step": 5071
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016073715045967588,
+      "loss": 0.705,
+      "step": 5072
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016072269900302918,
+      "loss": 0.7271,
+      "step": 5073
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00016070824553721637,
+      "loss": 0.6616,
+      "step": 5074
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016069379006271566,
+      "loss": 0.7037,
+      "step": 5075
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016067933258000542,
+      "loss": 0.7225,
+      "step": 5076
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016066487308956395,
+      "loss": 0.6553,
+      "step": 5077
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016065041159186967,
+      "loss": 0.8088,
+      "step": 5078
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016063594808740113,
+      "loss": 0.7518,
+      "step": 5079
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016062148257663686,
+      "loss": 0.727,
+      "step": 5080
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016060701506005547,
+      "loss": 0.7441,
+      "step": 5081
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016059254553813567,
+      "loss": 0.6884,
+      "step": 5082
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016057807401135624,
+      "loss": 0.706,
+      "step": 5083
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.000160563600480196,
+      "loss": 0.68,
+      "step": 5084
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016054912494513387,
+      "loss": 0.6836,
+      "step": 5085
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001605346474066487,
+      "loss": 0.6875,
+      "step": 5086
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016052016786521965,
+      "loss": 0.6808,
+      "step": 5087
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016050568632132575,
+      "loss": 0.7708,
+      "step": 5088
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016049120277544617,
+      "loss": 0.6545,
+      "step": 5089
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016047671722806012,
+      "loss": 0.7527,
+      "step": 5090
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001604622296796469,
+      "loss": 0.6664,
+      "step": 5091
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016044774013068588,
+      "loss": 0.7081,
+      "step": 5092
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016043324858165648,
+      "loss": 0.7379,
+      "step": 5093
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001604187550330382,
+      "loss": 0.6839,
+      "step": 5094
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016040425948531052,
+      "loss": 0.7928,
+      "step": 5095
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001603897619389532,
+      "loss": 0.6491,
+      "step": 5096
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001603752623944458,
+      "loss": 0.7441,
+      "step": 5097
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016036076085226814,
+      "loss": 0.7818,
+      "step": 5098
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016034625731290003,
+      "loss": 0.705,
+      "step": 5099
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016033175177682137,
+      "loss": 0.7438,
+      "step": 5100
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016031724424451208,
+      "loss": 0.7184,
+      "step": 5101
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001603027347164522,
+      "loss": 0.7135,
+      "step": 5102
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001602882231931218,
+      "loss": 0.8175,
+      "step": 5103
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016027370967500103,
+      "loss": 0.739,
+      "step": 5104
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016025919416257013,
+      "loss": 0.7385,
+      "step": 5105
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001602446766563094,
+      "loss": 0.781,
+      "step": 5106
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016023015715669914,
+      "loss": 0.7847,
+      "step": 5107
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001602156356642198,
+      "loss": 0.6864,
+      "step": 5108
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016020111217935183,
+      "loss": 0.7456,
+      "step": 5109
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016018658670257578,
+      "loss": 0.7489,
+      "step": 5110
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016017205923437227,
+      "loss": 0.7402,
+      "step": 5111
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016015752977522203,
+      "loss": 0.7529,
+      "step": 5112
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016014299832560575,
+      "loss": 0.7723,
+      "step": 5113
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016012846488600423,
+      "loss": 0.7186,
+      "step": 5114
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016011392945689833,
+      "loss": 0.7656,
+      "step": 5115
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001600993920387691,
+      "loss": 0.6987,
+      "step": 5116
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016008485263209742,
+      "loss": 0.6843,
+      "step": 5117
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00016007031123736444,
+      "loss": 0.7778,
+      "step": 5118
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00016005576785505126,
+      "loss": 0.7251,
+      "step": 5119
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00016004122248563913,
+      "loss": 0.7458,
+      "step": 5120
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00016002667512960927,
+      "loss": 0.7319,
+      "step": 5121
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00016001212578744306,
+      "loss": 0.7235,
+      "step": 5122
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015999757445962187,
+      "loss": 0.7495,
+      "step": 5123
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015998302114662716,
+      "loss": 0.6418,
+      "step": 5124
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001599684658489405,
+      "loss": 0.7002,
+      "step": 5125
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001599539085670435,
+      "loss": 0.6765,
+      "step": 5126
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015993934930141773,
+      "loss": 0.6905,
+      "step": 5127
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015992478805254505,
+      "loss": 0.6751,
+      "step": 5128
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015991022482090718,
+      "loss": 0.6603,
+      "step": 5129
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.000159895659606986,
+      "loss": 0.7158,
+      "step": 5130
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001598810924112634,
+      "loss": 0.6301,
+      "step": 5131
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015986652323422146,
+      "loss": 0.8029,
+      "step": 5132
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015985195207634213,
+      "loss": 0.7254,
+      "step": 5133
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001598373789381076,
+      "loss": 0.7298,
+      "step": 5134
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015982280382000007,
+      "loss": 0.7201,
+      "step": 5135
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001598082267225018,
+      "loss": 0.6937,
+      "step": 5136
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015979364764609504,
+      "loss": 0.7175,
+      "step": 5137
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015977906659126224,
+      "loss": 0.7247,
+      "step": 5138
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015976448355848585,
+      "loss": 0.7242,
+      "step": 5139
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015974989854824833,
+      "loss": 0.7248,
+      "step": 5140
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015973531156103233,
+      "loss": 0.7022,
+      "step": 5141
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015972072259732046,
+      "loss": 0.6791,
+      "step": 5142
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015970613165759544,
+      "loss": 0.6983,
+      "step": 5143
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015969153874234007,
+      "loss": 0.7425,
+      "step": 5144
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015967694385203714,
+      "loss": 0.6901,
+      "step": 5145
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015966234698716963,
+      "loss": 0.672,
+      "step": 5146
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015964774814822044,
+      "loss": 0.726,
+      "step": 5147
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001596331473356727,
+      "loss": 0.7205,
+      "step": 5148
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015961854455000943,
+      "loss": 0.6905,
+      "step": 5149
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015960393979171387,
+      "loss": 0.7092,
+      "step": 5150
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015958933306126918,
+      "loss": 0.7001,
+      "step": 5151
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015957472435915873,
+      "loss": 0.7004,
+      "step": 5152
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015956011368586584,
+      "loss": 0.6652,
+      "step": 5153
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015954550104187398,
+      "loss": 0.6514,
+      "step": 5154
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001595308864276666,
+      "loss": 0.6674,
+      "step": 5155
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015951626984372731,
+      "loss": 0.6928,
+      "step": 5156
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015950165129053972,
+      "loss": 0.7241,
+      "step": 5157
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001594870307685875,
+      "loss": 0.6749,
+      "step": 5158
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015947240827835448,
+      "loss": 0.7094,
+      "step": 5159
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015945778382032435,
+      "loss": 0.7237,
+      "step": 5160
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00015944315739498113,
+      "loss": 0.7806,
+      "step": 5161
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001594285290028087,
+      "loss": 0.6632,
+      "step": 5162
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001594138986442911,
+      "loss": 0.6319,
+      "step": 5163
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001593992663199124,
+      "loss": 0.8023,
+      "step": 5164
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001593846320301568,
+      "loss": 0.6772,
+      "step": 5165
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001593699957755084,
+      "loss": 0.752,
+      "step": 5166
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015935535755645162,
+      "loss": 0.6644,
+      "step": 5167
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015934071737347066,
+      "loss": 0.7653,
+      "step": 5168
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015932607522705007,
+      "loss": 0.7215,
+      "step": 5169
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015931143111767423,
+      "loss": 0.7191,
+      "step": 5170
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015929678504582765,
+      "loss": 0.6904,
+      "step": 5171
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015928213701199505,
+      "loss": 0.7283,
+      "step": 5172
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.000159267487016661,
+      "loss": 0.6525,
+      "step": 5173
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001592528350603103,
+      "loss": 0.6936,
+      "step": 5174
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015923818114342767,
+      "loss": 0.6603,
+      "step": 5175
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015922352526649803,
+      "loss": 0.6929,
+      "step": 5176
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015920886743000628,
+      "loss": 0.7357,
+      "step": 5177
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015919420763443743,
+      "loss": 0.6748,
+      "step": 5178
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015917954588027652,
+      "loss": 0.7238,
+      "step": 5179
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001591648821680087,
+      "loss": 0.7297,
+      "step": 5180
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015915021649811912,
+      "loss": 0.729,
+      "step": 5181
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015913554887109305,
+      "loss": 0.7022,
+      "step": 5182
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001591208792874158,
+      "loss": 0.6981,
+      "step": 5183
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015910620774757278,
+      "loss": 0.7613,
+      "step": 5184
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001590915342520494,
+      "loss": 0.6751,
+      "step": 5185
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001590768588013312,
+      "loss": 0.6573,
+      "step": 5186
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001590621813959037,
+      "loss": 0.6846,
+      "step": 5187
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001590475020362526,
+      "loss": 0.6527,
+      "step": 5188
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001590328207228636,
+      "loss": 0.6752,
+      "step": 5189
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015901813745622243,
+      "loss": 0.6898,
+      "step": 5190
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015900345223681495,
+      "loss": 0.6861,
+      "step": 5191
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015898876506512704,
+      "loss": 0.6985,
+      "step": 5192
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015897407594164467,
+      "loss": 0.6646,
+      "step": 5193
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015895938486685385,
+      "loss": 0.753,
+      "step": 5194
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001589446918412407,
+      "loss": 0.7444,
+      "step": 5195
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001589299968652914,
+      "loss": 0.7217,
+      "step": 5196
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015891529993949214,
+      "loss": 0.7155,
+      "step": 5197
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0001589006010643292,
+      "loss": 0.7103,
+      "step": 5198
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015888590024028894,
+      "loss": 0.6981,
+      "step": 5199
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015887119746785776,
+      "loss": 0.7209,
+      "step": 5200
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015885649274752215,
+      "loss": 0.767,
+      "step": 5201
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015884178607976864,
+      "loss": 0.6802,
+      "step": 5202
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015882707746508388,
+      "loss": 0.8035,
+      "step": 5203
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00015881236690395447,
+      "loss": 0.7089,
+      "step": 5204
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015879765439686722,
+      "loss": 0.6592,
+      "step": 5205
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015878293994430888,
+      "loss": 0.6789,
+      "step": 5206
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015876822354676638,
+      "loss": 0.7403,
+      "step": 5207
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015875350520472655,
+      "loss": 0.7286,
+      "step": 5208
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015873878491867645,
+      "loss": 0.702,
+      "step": 5209
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015872406268910316,
+      "loss": 0.8267,
+      "step": 5210
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015870933851649374,
+      "loss": 0.6578,
+      "step": 5211
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001586946124013354,
+      "loss": 0.6901,
+      "step": 5212
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015867988434411544,
+      "loss": 0.7255,
+      "step": 5213
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001586651543453211,
+      "loss": 0.7437,
+      "step": 5214
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015865042240543984,
+      "loss": 0.7497,
+      "step": 5215
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.000158635688524959,
+      "loss": 0.6637,
+      "step": 5216
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015862095270436616,
+      "loss": 0.7227,
+      "step": 5217
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015860621494414892,
+      "loss": 0.6801,
+      "step": 5218
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015859147524479483,
+      "loss": 0.7069,
+      "step": 5219
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015857673360679164,
+      "loss": 0.6542,
+      "step": 5220
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001585619900306271,
+      "loss": 0.6805,
+      "step": 5221
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015854724451678907,
+      "loss": 0.742,
+      "step": 5222
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015853249706576544,
+      "loss": 0.731,
+      "step": 5223
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015851774767804413,
+      "loss": 0.7773,
+      "step": 5224
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015850299635411315,
+      "loss": 0.6978,
+      "step": 5225
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015848824309446066,
+      "loss": 0.6903,
+      "step": 5226
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015847348789957476,
+      "loss": 0.6866,
+      "step": 5227
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015845873076994366,
+      "loss": 0.6822,
+      "step": 5228
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015844397170605565,
+      "loss": 0.6712,
+      "step": 5229
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015842921070839908,
+      "loss": 0.7153,
+      "step": 5230
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001584144477774623,
+      "loss": 0.7299,
+      "step": 5231
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015839968291373385,
+      "loss": 0.6926,
+      "step": 5232
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015838491611770225,
+      "loss": 0.6324,
+      "step": 5233
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001583701473898561,
+      "loss": 0.7838,
+      "step": 5234
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.000158355376730684,
+      "loss": 0.6941,
+      "step": 5235
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015834060414067472,
+      "loss": 0.7214,
+      "step": 5236
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001583258296203171,
+      "loss": 0.7165,
+      "step": 5237
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001583110531700999,
+      "loss": 0.7171,
+      "step": 5238
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015829627479051209,
+      "loss": 0.7456,
+      "step": 5239
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015828149448204263,
+      "loss": 0.7388,
+      "step": 5240
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015826671224518056,
+      "loss": 0.7085,
+      "step": 5241
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015825192808041502,
+      "loss": 0.6297,
+      "step": 5242
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015823714198823517,
+      "loss": 0.7963,
+      "step": 5243
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015822235396913023,
+      "loss": 0.7147,
+      "step": 5244
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015820756402358952,
+      "loss": 0.6649,
+      "step": 5245
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001581927721521024,
+      "loss": 0.7713,
+      "step": 5246
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00015817797835515827,
+      "loss": 0.6672,
+      "step": 5247
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015816318263324664,
+      "loss": 0.6915,
+      "step": 5248
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015814838498685705,
+      "loss": 0.7113,
+      "step": 5249
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015813358541647915,
+      "loss": 0.6665,
+      "step": 5250
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015811878392260265,
+      "loss": 0.7028,
+      "step": 5251
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001581039805057172,
+      "loss": 0.6986,
+      "step": 5252
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015808917516631262,
+      "loss": 0.7754,
+      "step": 5253
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001580743679048789,
+      "loss": 0.6665,
+      "step": 5254
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015805955872190584,
+      "loss": 0.7124,
+      "step": 5255
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015804474761788351,
+      "loss": 0.6771,
+      "step": 5256
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.000158029934593302,
+      "loss": 0.724,
+      "step": 5257
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015801511964865135,
+      "loss": 0.7464,
+      "step": 5258
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001580003027844218,
+      "loss": 0.7309,
+      "step": 5259
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001579854840011036,
+      "loss": 0.7453,
+      "step": 5260
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001579706632991871,
+      "loss": 0.7762,
+      "step": 5261
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015795584067916264,
+      "loss": 0.6872,
+      "step": 5262
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015794101614152067,
+      "loss": 0.7001,
+      "step": 5263
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001579261896867517,
+      "loss": 0.6885,
+      "step": 5264
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015791136131534627,
+      "loss": 0.7172,
+      "step": 5265
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015789653102779512,
+      "loss": 0.7555,
+      "step": 5266
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015788169882458881,
+      "loss": 0.642,
+      "step": 5267
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015786686470621822,
+      "loss": 0.6579,
+      "step": 5268
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015785202867317407,
+      "loss": 0.6197,
+      "step": 5269
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015783719072594732,
+      "loss": 0.6838,
+      "step": 5270
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015782235086502893,
+      "loss": 0.747,
+      "step": 5271
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015780750909090988,
+      "loss": 0.7075,
+      "step": 5272
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001577926654040812,
+      "loss": 0.6606,
+      "step": 5273
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001577778198050341,
+      "loss": 0.6944,
+      "step": 5274
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015776297229425982,
+      "loss": 0.7017,
+      "step": 5275
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015774812287224955,
+      "loss": 0.6453,
+      "step": 5276
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015773327153949465,
+      "loss": 0.7336,
+      "step": 5277
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015771841829648653,
+      "loss": 0.746,
+      "step": 5278
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001577035631437166,
+      "loss": 0.7513,
+      "step": 5279
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015768870608167643,
+      "loss": 0.7039,
+      "step": 5280
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001576738471108576,
+      "loss": 0.7422,
+      "step": 5281
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015765898623175176,
+      "loss": 0.6751,
+      "step": 5282
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015764412344485056,
+      "loss": 0.7412,
+      "step": 5283
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015762925875064583,
+      "loss": 0.6919,
+      "step": 5284
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001576143921496294,
+      "loss": 0.6723,
+      "step": 5285
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015759952364229318,
+      "loss": 0.7097,
+      "step": 5286
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001575846532291291,
+      "loss": 0.6936,
+      "step": 5287
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001575697809106292,
+      "loss": 0.6982,
+      "step": 5288
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015755490668728557,
+      "loss": 0.6733,
+      "step": 5289
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00015754003055959035,
+      "loss": 0.7124,
+      "step": 5290
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015752515252803577,
+      "loss": 0.6689,
+      "step": 5291
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015751027259311413,
+      "loss": 0.685,
+      "step": 5292
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015749539075531773,
+      "loss": 0.6766,
+      "step": 5293
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.000157480507015139,
+      "loss": 0.6956,
+      "step": 5294
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001574656213730704,
+      "loss": 0.7139,
+      "step": 5295
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015745073382960443,
+      "loss": 0.7512,
+      "step": 5296
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015743584438523373,
+      "loss": 0.7031,
+      "step": 5297
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015742095304045094,
+      "loss": 0.7144,
+      "step": 5298
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015740605979574877,
+      "loss": 0.6895,
+      "step": 5299
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015739116465161997,
+      "loss": 0.7007,
+      "step": 5300
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015737626760855747,
+      "loss": 0.677,
+      "step": 5301
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015736136866705407,
+      "loss": 0.6635,
+      "step": 5302
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015734646782760282,
+      "loss": 0.6811,
+      "step": 5303
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015733156509069673,
+      "loss": 0.696,
+      "step": 5304
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015731666045682889,
+      "loss": 0.6845,
+      "step": 5305
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015730175392649243,
+      "loss": 0.7144,
+      "step": 5306
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015728684550018064,
+      "loss": 0.7704,
+      "step": 5307
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015727193517838678,
+      "loss": 0.716,
+      "step": 5308
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015725702296160416,
+      "loss": 0.7301,
+      "step": 5309
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001572421088503262,
+      "loss": 0.7065,
+      "step": 5310
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015722719284504637,
+      "loss": 0.7663,
+      "step": 5311
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001572122749462582,
+      "loss": 0.6689,
+      "step": 5312
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015719735515445536,
+      "loss": 0.7112,
+      "step": 5313
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001571824334701314,
+      "loss": 0.7452,
+      "step": 5314
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015716750989378013,
+      "loss": 0.6958,
+      "step": 5315
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001571525844258953,
+      "loss": 0.6624,
+      "step": 5316
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001571376570669707,
+      "loss": 0.6827,
+      "step": 5317
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015712272781750035,
+      "loss": 0.787,
+      "step": 5318
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015710779667797814,
+      "loss": 0.719,
+      "step": 5319
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015709286364889814,
+      "loss": 0.6357,
+      "step": 5320
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015707792873075445,
+      "loss": 0.7439,
+      "step": 5321
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015706299192404122,
+      "loss": 0.7221,
+      "step": 5322
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001570480532292527,
+      "loss": 0.6704,
+      "step": 5323
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015703311264688309,
+      "loss": 0.75,
+      "step": 5324
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015701817017742684,
+      "loss": 0.7816,
+      "step": 5325
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015700322582137827,
+      "loss": 0.7113,
+      "step": 5326
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015698827957923194,
+      "loss": 0.704,
+      "step": 5327
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001569733314514823,
+      "loss": 0.6707,
+      "step": 5328
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015695838143862405,
+      "loss": 0.704,
+      "step": 5329
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015694342954115174,
+      "loss": 0.6711,
+      "step": 5330
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015692847575956016,
+      "loss": 0.7605,
+      "step": 5331
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001569135200943441,
+      "loss": 0.7233,
+      "step": 5332
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015689856254599832,
+      "loss": 0.7543,
+      "step": 5333
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00015688360311501783,
+      "loss": 0.6921,
+      "step": 5334
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015686864180189757,
+      "loss": 0.7464,
+      "step": 5335
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015685367860713255,
+      "loss": 0.7362,
+      "step": 5336
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015683871353121788,
+      "loss": 0.7159,
+      "step": 5337
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001568237465746487,
+      "loss": 0.7613,
+      "step": 5338
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001568087777379203,
+      "loss": 0.6609,
+      "step": 5339
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015679380702152788,
+      "loss": 0.6487,
+      "step": 5340
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015677883442596682,
+      "loss": 0.6534,
+      "step": 5341
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015676385995173252,
+      "loss": 0.7037,
+      "step": 5342
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015674888359932043,
+      "loss": 0.6933,
+      "step": 5343
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015673390536922611,
+      "loss": 0.7115,
+      "step": 5344
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015671892526194516,
+      "loss": 0.7013,
+      "step": 5345
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015670394327797323,
+      "loss": 0.7082,
+      "step": 5346
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.000156688959417806,
+      "loss": 0.648,
+      "step": 5347
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015667397368193932,
+      "loss": 0.7009,
+      "step": 5348
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015665898607086892,
+      "loss": 0.6999,
+      "step": 5349
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015664399658509083,
+      "loss": 0.7247,
+      "step": 5350
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015662900522510096,
+      "loss": 0.7366,
+      "step": 5351
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001566140119913953,
+      "loss": 0.6563,
+      "step": 5352
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015659901688446997,
+      "loss": 0.7142,
+      "step": 5353
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015658401990482118,
+      "loss": 0.7295,
+      "step": 5354
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015656902105294504,
+      "loss": 0.7254,
+      "step": 5355
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001565540203293379,
+      "loss": 0.7472,
+      "step": 5356
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015653901773449605,
+      "loss": 0.6438,
+      "step": 5357
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015652401326891595,
+      "loss": 0.6673,
+      "step": 5358
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015650900693309398,
+      "loss": 0.7681,
+      "step": 5359
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001564939987275267,
+      "loss": 0.6815,
+      "step": 5360
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015647898865271077,
+      "loss": 0.7082,
+      "step": 5361
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001564639767091427,
+      "loss": 0.721,
+      "step": 5362
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015644896289731928,
+      "loss": 0.6493,
+      "step": 5363
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001564339472177373,
+      "loss": 0.7687,
+      "step": 5364
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015641892967089354,
+      "loss": 0.6792,
+      "step": 5365
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015640391025728488,
+      "loss": 0.6657,
+      "step": 5366
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015638888897740838,
+      "loss": 0.6801,
+      "step": 5367
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015637386583176093,
+      "loss": 0.6697,
+      "step": 5368
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015635884082083968,
+      "loss": 0.7151,
+      "step": 5369
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015634381394514176,
+      "loss": 0.7314,
+      "step": 5370
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015632878520516436,
+      "loss": 0.723,
+      "step": 5371
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015631375460140476,
+      "loss": 0.7339,
+      "step": 5372
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015629872213436028,
+      "loss": 0.7379,
+      "step": 5373
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015628368780452832,
+      "loss": 0.7265,
+      "step": 5374
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001562686516124063,
+      "loss": 0.7533,
+      "step": 5375
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015625361355849177,
+      "loss": 0.7242,
+      "step": 5376
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00015623857364328227,
+      "loss": 0.6795,
+      "step": 5377
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015622353186727544,
+      "loss": 0.711,
+      "step": 5378
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015620848823096897,
+      "loss": 0.7575,
+      "step": 5379
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015619344273486063,
+      "loss": 0.696,
+      "step": 5380
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015617839537944824,
+      "loss": 0.7155,
+      "step": 5381
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001561633461652297,
+      "loss": 0.7998,
+      "step": 5382
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001561482950927029,
+      "loss": 0.7523,
+      "step": 5383
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001561332421623659,
+      "loss": 0.7423,
+      "step": 5384
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001561181873747167,
+      "loss": 0.6893,
+      "step": 5385
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015610313073025347,
+      "loss": 0.732,
+      "step": 5386
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001560880722294744,
+      "loss": 0.7088,
+      "step": 5387
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015607301187287774,
+      "loss": 0.7274,
+      "step": 5388
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015605794966096177,
+      "loss": 0.6843,
+      "step": 5389
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015604288559422491,
+      "loss": 0.6603,
+      "step": 5390
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015602781967316554,
+      "loss": 0.6858,
+      "step": 5391
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015601275189828219,
+      "loss": 0.769,
+      "step": 5392
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001559976822700734,
+      "loss": 0.6851,
+      "step": 5393
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001559826107890378,
+      "loss": 0.6809,
+      "step": 5394
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015596753745567406,
+      "loss": 0.7161,
+      "step": 5395
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001559524622704809,
+      "loss": 0.6185,
+      "step": 5396
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015593738523395716,
+      "loss": 0.6461,
+      "step": 5397
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015592230634660167,
+      "loss": 0.7389,
+      "step": 5398
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001559072256089134,
+      "loss": 0.7177,
+      "step": 5399
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001558921430213913,
+      "loss": 0.7615,
+      "step": 5400
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001558770585845344,
+      "loss": 0.745,
+      "step": 5401
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015586197229884184,
+      "loss": 0.781,
+      "step": 5402
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015584688416481277,
+      "loss": 0.6849,
+      "step": 5403
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015583179418294642,
+      "loss": 0.7185,
+      "step": 5404
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015581670235374211,
+      "loss": 0.7088,
+      "step": 5405
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015580160867769913,
+      "loss": 0.7041,
+      "step": 5406
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015578651315531695,
+      "loss": 0.68,
+      "step": 5407
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.000155771415787095,
+      "loss": 0.7677,
+      "step": 5408
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015575631657353288,
+      "loss": 0.6686,
+      "step": 5409
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015574121551513011,
+      "loss": 0.7174,
+      "step": 5410
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015572611261238642,
+      "loss": 0.7535,
+      "step": 5411
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015571100786580146,
+      "loss": 0.6805,
+      "step": 5412
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015569590127587505,
+      "loss": 0.7128,
+      "step": 5413
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015568079284310703,
+      "loss": 0.7669,
+      "step": 5414
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001556656825679973,
+      "loss": 0.7198,
+      "step": 5415
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001556505704510458,
+      "loss": 0.652,
+      "step": 5416
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015563545649275256,
+      "loss": 0.6928,
+      "step": 5417
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015562034069361766,
+      "loss": 0.6619,
+      "step": 5418
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00015560522305414125,
+      "loss": 0.6715,
+      "step": 5419
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001555901035748236,
+      "loss": 0.6871,
+      "step": 5420
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015557498225616487,
+      "loss": 0.7593,
+      "step": 5421
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015555985909866544,
+      "loss": 0.6912,
+      "step": 5422
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001555447341028257,
+      "loss": 0.716,
+      "step": 5423
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001555296072691461,
+      "loss": 0.7232,
+      "step": 5424
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015551447859812716,
+      "loss": 0.6873,
+      "step": 5425
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015549934809026941,
+      "loss": 0.6761,
+      "step": 5426
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001554842157460735,
+      "loss": 0.6692,
+      "step": 5427
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015546908156604015,
+      "loss": 0.6214,
+      "step": 5428
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015545394555067005,
+      "loss": 0.7082,
+      "step": 5429
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001554388077004641,
+      "loss": 0.7539,
+      "step": 5430
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015542366801592311,
+      "loss": 0.7316,
+      "step": 5431
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015540852649754806,
+      "loss": 0.7609,
+      "step": 5432
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015539338314583988,
+      "loss": 0.6957,
+      "step": 5433
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015537823796129971,
+      "loss": 0.7272,
+      "step": 5434
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015536309094442863,
+      "loss": 0.7821,
+      "step": 5435
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015534794209572778,
+      "loss": 0.6709,
+      "step": 5436
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015533279141569843,
+      "loss": 0.7704,
+      "step": 5437
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001553176389048419,
+      "loss": 0.703,
+      "step": 5438
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015530248456365953,
+      "loss": 0.6875,
+      "step": 5439
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015528732839265272,
+      "loss": 0.7356,
+      "step": 5440
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.000155272170392323,
+      "loss": 0.7078,
+      "step": 5441
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015525701056317187,
+      "loss": 0.6774,
+      "step": 5442
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015524184890570094,
+      "loss": 0.7454,
+      "step": 5443
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015522668542041187,
+      "loss": 0.6794,
+      "step": 5444
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015521152010780642,
+      "loss": 0.7714,
+      "step": 5445
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015519635296838634,
+      "loss": 0.6947,
+      "step": 5446
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015518118400265345,
+      "loss": 0.7266,
+      "step": 5447
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015516601321110968,
+      "loss": 0.7108,
+      "step": 5448
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.000155150840594257,
+      "loss": 0.7489,
+      "step": 5449
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015513566615259742,
+      "loss": 0.7167,
+      "step": 5450
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015512048988663307,
+      "loss": 0.6687,
+      "step": 5451
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015510531179686602,
+      "loss": 0.6977,
+      "step": 5452
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015509013188379854,
+      "loss": 0.7126,
+      "step": 5453
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001550749501479329,
+      "loss": 0.7436,
+      "step": 5454
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015505976658977138,
+      "loss": 0.5818,
+      "step": 5455
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015504458120981632,
+      "loss": 0.6792,
+      "step": 5456
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015502939400857034,
+      "loss": 0.6662,
+      "step": 5457
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015501420498653578,
+      "loss": 0.6232,
+      "step": 5458
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001549990141442153,
+      "loss": 0.6668,
+      "step": 5459
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015498382148211145,
+      "loss": 0.7566,
+      "step": 5460
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015496862700072698,
+      "loss": 0.729,
+      "step": 5461
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015495343070056464,
+      "loss": 0.7556,
+      "step": 5462
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00015493823258212722,
+      "loss": 0.7765,
+      "step": 5463
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001549230326459176,
+      "loss": 0.7266,
+      "step": 5464
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001549078308924387,
+      "loss": 0.6456,
+      "step": 5465
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015489262732219352,
+      "loss": 0.7222,
+      "step": 5466
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015487742193568503,
+      "loss": 0.681,
+      "step": 5467
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015486221473341648,
+      "loss": 0.6887,
+      "step": 5468
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015484700571589095,
+      "loss": 0.7154,
+      "step": 5469
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015483179488361167,
+      "loss": 0.6497,
+      "step": 5470
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015481658223708196,
+      "loss": 0.6862,
+      "step": 5471
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015480136777680514,
+      "loss": 0.7414,
+      "step": 5472
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015478615150328463,
+      "loss": 0.6986,
+      "step": 5473
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015477093341702394,
+      "loss": 0.7183,
+      "step": 5474
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001547557135185265,
+      "loss": 0.6895,
+      "step": 5475
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015474049180829602,
+      "loss": 0.7242,
+      "step": 5476
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015472526828683603,
+      "loss": 0.6697,
+      "step": 5477
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015471004295465035,
+      "loss": 0.7585,
+      "step": 5478
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015469481581224272,
+      "loss": 0.7699,
+      "step": 5479
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015467958686011689,
+      "loss": 0.7497,
+      "step": 5480
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015466435609877682,
+      "loss": 0.6666,
+      "step": 5481
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015464912352872648,
+      "loss": 0.6701,
+      "step": 5482
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015463388915046983,
+      "loss": 0.7171,
+      "step": 5483
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015461865296451094,
+      "loss": 0.6837,
+      "step": 5484
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015460341497135395,
+      "loss": 0.7132,
+      "step": 5485
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015458817517150305,
+      "loss": 0.6877,
+      "step": 5486
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015457293356546252,
+      "loss": 0.7375,
+      "step": 5487
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001545576901537366,
+      "loss": 0.7164,
+      "step": 5488
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001545424449368297,
+      "loss": 0.753,
+      "step": 5489
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015452719791524624,
+      "loss": 0.6215,
+      "step": 5490
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015451194908949073,
+      "loss": 0.7405,
+      "step": 5491
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015449669846006765,
+      "loss": 0.6539,
+      "step": 5492
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015449669846006765,
+      "loss": 0.7083,
+      "step": 5493
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015448144602748167,
+      "loss": 0.7882,
+      "step": 5494
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015446619179223741,
+      "loss": 0.657,
+      "step": 5495
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015445093575483968,
+      "loss": 0.7455,
+      "step": 5496
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015443567791579316,
+      "loss": 0.7543,
+      "step": 5497
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015442041827560274,
+      "loss": 0.7413,
+      "step": 5498
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015440515683477334,
+      "loss": 0.7048,
+      "step": 5499
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001543898935938099,
+      "loss": 0.6354,
+      "step": 5500
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015437462855321746,
+      "loss": 0.7136,
+      "step": 5501
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015435936171350107,
+      "loss": 0.6734,
+      "step": 5502
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015434409307516593,
+      "loss": 0.6775,
+      "step": 5503
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015432882263871717,
+      "loss": 0.701,
+      "step": 5504
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015431355040466013,
+      "loss": 0.6759,
+      "step": 5505
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00015429827637350007,
+      "loss": 0.6878,
+      "step": 5506
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001542830005457424,
+      "loss": 0.6851,
+      "step": 5507
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015426772292189255,
+      "loss": 0.7414,
+      "step": 5508
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015425244350245602,
+      "loss": 0.7072,
+      "step": 5509
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015423716228793834,
+      "loss": 0.652,
+      "step": 5510
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015422187927884525,
+      "loss": 0.6775,
+      "step": 5511
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015420659447568222,
+      "loss": 0.736,
+      "step": 5512
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015419130787895516,
+      "loss": 0.6885,
+      "step": 5513
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001541760194891698,
+      "loss": 0.7175,
+      "step": 5514
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015416072930683202,
+      "loss": 0.7097,
+      "step": 5515
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015414543733244772,
+      "loss": 0.6895,
+      "step": 5516
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015413014356652286,
+      "loss": 0.7429,
+      "step": 5517
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015411484800956348,
+      "loss": 0.7122,
+      "step": 5518
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001540995506620757,
+      "loss": 0.6934,
+      "step": 5519
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015408425152456563,
+      "loss": 0.6821,
+      "step": 5520
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001540689505975395,
+      "loss": 0.6946,
+      "step": 5521
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001540536478815036,
+      "loss": 0.7108,
+      "step": 5522
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015403834337696424,
+      "loss": 0.6905,
+      "step": 5523
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015402303708442783,
+      "loss": 0.7327,
+      "step": 5524
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015400772900440077,
+      "loss": 0.7169,
+      "step": 5525
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001539924191373896,
+      "loss": 0.6602,
+      "step": 5526
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001539771074839009,
+      "loss": 0.6371,
+      "step": 5527
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015396179404444124,
+      "loss": 0.741,
+      "step": 5528
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001539464788195174,
+      "loss": 0.6989,
+      "step": 5529
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.000153931161809636,
+      "loss": 0.7565,
+      "step": 5530
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015391584301530396,
+      "loss": 0.7506,
+      "step": 5531
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015390052243702805,
+      "loss": 0.6911,
+      "step": 5532
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015388520007531524,
+      "loss": 0.7319,
+      "step": 5533
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015386987593067253,
+      "loss": 0.6693,
+      "step": 5534
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001538545500036069,
+      "loss": 0.7156,
+      "step": 5535
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015383922229462549,
+      "loss": 0.7084,
+      "step": 5536
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015382389280423542,
+      "loss": 0.6447,
+      "step": 5537
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015380856153294397,
+      "loss": 0.6903,
+      "step": 5538
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015379322848125832,
+      "loss": 0.7594,
+      "step": 5539
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015377789364968587,
+      "loss": 0.7209,
+      "step": 5540
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015376255703873404,
+      "loss": 0.6788,
+      "step": 5541
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015374721864891018,
+      "loss": 0.7116,
+      "step": 5542
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015373187848072192,
+      "loss": 0.7663,
+      "step": 5543
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001537165365346767,
+      "loss": 0.6818,
+      "step": 5544
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015370119281128228,
+      "loss": 0.6723,
+      "step": 5545
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015368584731104625,
+      "loss": 0.6938,
+      "step": 5546
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015367050003447637,
+      "loss": 0.6852,
+      "step": 5547
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001536551509820805,
+      "loss": 0.6399,
+      "step": 5548
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015363980015436644,
+      "loss": 0.69,
+      "step": 5549
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00015362444755184212,
+      "loss": 0.6819,
+      "step": 5550
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015360909317501555,
+      "loss": 0.7332,
+      "step": 5551
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015359373702439477,
+      "loss": 0.6889,
+      "step": 5552
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015357837910048784,
+      "loss": 0.6694,
+      "step": 5553
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015356301940380298,
+      "loss": 0.7817,
+      "step": 5554
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015354765793484834,
+      "loss": 0.7484,
+      "step": 5555
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015353229469413217,
+      "loss": 0.751,
+      "step": 5556
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001535169296821629,
+      "loss": 0.673,
+      "step": 5557
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015350156289944885,
+      "loss": 0.7204,
+      "step": 5558
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015348619434649852,
+      "loss": 0.7315,
+      "step": 5559
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015347082402382035,
+      "loss": 0.6864,
+      "step": 5560
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015345545193192297,
+      "loss": 0.717,
+      "step": 5561
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015344007807131495,
+      "loss": 0.6874,
+      "step": 5562
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015342470244250503,
+      "loss": 0.7277,
+      "step": 5563
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015340932504600188,
+      "loss": 0.7692,
+      "step": 5564
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015339394588231438,
+      "loss": 0.7395,
+      "step": 5565
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015337856495195137,
+      "loss": 0.7304,
+      "step": 5566
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001533631822554217,
+      "loss": 0.7775,
+      "step": 5567
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015334779779323444,
+      "loss": 0.6936,
+      "step": 5568
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015333241156589855,
+      "loss": 0.6473,
+      "step": 5569
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015331702357392318,
+      "loss": 0.6723,
+      "step": 5570
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015330163381781749,
+      "loss": 0.6929,
+      "step": 5571
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001532862422980906,
+      "loss": 0.7466,
+      "step": 5572
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015327084901525182,
+      "loss": 0.7057,
+      "step": 5573
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001532554539698105,
+      "loss": 0.6947,
+      "step": 5574
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015324005716227607,
+      "loss": 0.713,
+      "step": 5575
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015322465859315786,
+      "loss": 0.7134,
+      "step": 5576
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015320925826296545,
+      "loss": 0.718,
+      "step": 5577
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015319385617220836,
+      "loss": 0.6731,
+      "step": 5578
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001531784523213962,
+      "loss": 0.6967,
+      "step": 5579
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001531630467110387,
+      "loss": 0.6893,
+      "step": 5580
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001531476393416456,
+      "loss": 0.7108,
+      "step": 5581
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015313223021372663,
+      "loss": 0.7128,
+      "step": 5582
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015311681932779164,
+      "loss": 0.7618,
+      "step": 5583
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015310140668435058,
+      "loss": 0.6977,
+      "step": 5584
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015308599228391342,
+      "loss": 0.6991,
+      "step": 5585
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001530705761269902,
+      "loss": 0.6616,
+      "step": 5586
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015305515821409092,
+      "loss": 0.6693,
+      "step": 5587
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015303973854572577,
+      "loss": 0.676,
+      "step": 5588
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015302431712240497,
+      "loss": 0.6724,
+      "step": 5589
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015300889394463871,
+      "loss": 0.6504,
+      "step": 5590
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015299346901293743,
+      "loss": 0.7938,
+      "step": 5591
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015297804232781138,
+      "loss": 0.662,
+      "step": 5592
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00015296261388977108,
+      "loss": 0.74,
+      "step": 5593
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015294718369932692,
+      "loss": 0.7233,
+      "step": 5594
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015293175175698953,
+      "loss": 0.64,
+      "step": 5595
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015291631806326947,
+      "loss": 0.6816,
+      "step": 5596
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015290088261867748,
+      "loss": 0.7381,
+      "step": 5597
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001528854454237242,
+      "loss": 0.6644,
+      "step": 5598
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001528700064789204,
+      "loss": 0.6661,
+      "step": 5599
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.000152854565784777,
+      "loss": 0.6199,
+      "step": 5600
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001528391233418048,
+      "loss": 0.7011,
+      "step": 5601
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015282367915051486,
+      "loss": 0.7196,
+      "step": 5602
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015280823321141808,
+      "loss": 0.6804,
+      "step": 5603
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001527927855250256,
+      "loss": 0.6653,
+      "step": 5604
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015277733609184852,
+      "loss": 0.7461,
+      "step": 5605
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015276188491239803,
+      "loss": 0.6657,
+      "step": 5606
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015274643198718538,
+      "loss": 0.7566,
+      "step": 5607
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001527309773167218,
+      "loss": 0.689,
+      "step": 5608
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001527155209015188,
+      "loss": 0.7067,
+      "step": 5609
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015270006274208765,
+      "loss": 0.667,
+      "step": 5610
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001526846028389399,
+      "loss": 0.6683,
+      "step": 5611
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.000152669141192587,
+      "loss": 0.6714,
+      "step": 5612
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015265367780354066,
+      "loss": 0.7001,
+      "step": 5613
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015263821267231243,
+      "loss": 0.6963,
+      "step": 5614
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015262274579941405,
+      "loss": 0.7578,
+      "step": 5615
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015260727718535727,
+      "loss": 0.7479,
+      "step": 5616
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001525918068306539,
+      "loss": 0.6954,
+      "step": 5617
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015257633473581583,
+      "loss": 0.6623,
+      "step": 5618
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015256086090135505,
+      "loss": 0.6434,
+      "step": 5619
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015254538532778343,
+      "loss": 0.6838,
+      "step": 5620
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001525299080156131,
+      "loss": 0.7002,
+      "step": 5621
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015251442896535618,
+      "loss": 0.6826,
+      "step": 5622
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015249894817752477,
+      "loss": 0.7232,
+      "step": 5623
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015248346565263114,
+      "loss": 0.7384,
+      "step": 5624
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015246798139118757,
+      "loss": 0.7251,
+      "step": 5625
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015245249539370636,
+      "loss": 0.7295,
+      "step": 5626
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015243700766069993,
+      "loss": 0.6879,
+      "step": 5627
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015242151819268074,
+      "loss": 0.7053,
+      "step": 5628
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015240602699016128,
+      "loss": 0.6642,
+      "step": 5629
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015239053405365408,
+      "loss": 0.7318,
+      "step": 5630
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015237503938367186,
+      "loss": 0.7483,
+      "step": 5631
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015235954298072724,
+      "loss": 0.682,
+      "step": 5632
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015234404484533294,
+      "loss": 0.7201,
+      "step": 5633
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015232854497800177,
+      "loss": 0.67,
+      "step": 5634
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015231304337924662,
+      "loss": 0.6761,
+      "step": 5635
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00015229754004958038,
+      "loss": 0.6893,
+      "step": 5636
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015228203498951598,
+      "loss": 0.6891,
+      "step": 5637
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001522665281995665,
+      "loss": 0.6751,
+      "step": 5638
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015225101968024504,
+      "loss": 0.6895,
+      "step": 5639
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015223550943206462,
+      "loss": 0.6154,
+      "step": 5640
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015221999745553854,
+      "loss": 0.7101,
+      "step": 5641
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015220448375118,
+      "loss": 0.7441,
+      "step": 5642
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015218896831950237,
+      "loss": 0.7386,
+      "step": 5643
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015217345116101902,
+      "loss": 0.7068,
+      "step": 5644
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015215793227624328,
+      "loss": 0.6744,
+      "step": 5645
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001521424116656887,
+      "loss": 0.7385,
+      "step": 5646
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001521268893298688,
+      "loss": 0.7022,
+      "step": 5647
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001521113652692972,
+      "loss": 0.7673,
+      "step": 5648
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015209583948448753,
+      "loss": 0.6874,
+      "step": 5649
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015208031197595356,
+      "loss": 0.7132,
+      "step": 5650
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015206478274420895,
+      "loss": 0.6985,
+      "step": 5651
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001520492517897676,
+      "loss": 0.7084,
+      "step": 5652
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001520337191131434,
+      "loss": 0.6884,
+      "step": 5653
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001520181847148502,
+      "loss": 0.6769,
+      "step": 5654
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015200264859540214,
+      "loss": 0.6944,
+      "step": 5655
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015198711075531314,
+      "loss": 0.7186,
+      "step": 5656
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015197157119509736,
+      "loss": 0.7517,
+      "step": 5657
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015195602991526896,
+      "loss": 0.6427,
+      "step": 5658
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001519404869163422,
+      "loss": 0.7075,
+      "step": 5659
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001519249421988313,
+      "loss": 0.7226,
+      "step": 5660
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015190939576325066,
+      "loss": 0.6844,
+      "step": 5661
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015189384761011461,
+      "loss": 0.7503,
+      "step": 5662
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015187829773993762,
+      "loss": 0.6064,
+      "step": 5663
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015186274615323417,
+      "loss": 0.7497,
+      "step": 5664
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001518471928505189,
+      "loss": 0.6359,
+      "step": 5665
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015183163783230645,
+      "loss": 0.7164,
+      "step": 5666
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015181608109911138,
+      "loss": 0.7967,
+      "step": 5667
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015180052265144846,
+      "loss": 0.6336,
+      "step": 5668
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015178496248983254,
+      "loss": 0.7086,
+      "step": 5669
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015176940061477841,
+      "loss": 0.7549,
+      "step": 5670
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015175383702680106,
+      "loss": 0.6785,
+      "step": 5671
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015173827172641535,
+      "loss": 0.6738,
+      "step": 5672
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015172270471413632,
+      "loss": 0.6379,
+      "step": 5673
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015170713599047905,
+      "loss": 0.7746,
+      "step": 5674
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001516915655559587,
+      "loss": 0.6663,
+      "step": 5675
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015167599341109046,
+      "loss": 0.667,
+      "step": 5676
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015166041955638954,
+      "loss": 0.7281,
+      "step": 5677
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015164484399237123,
+      "loss": 0.621,
+      "step": 5678
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00015162926671955093,
+      "loss": 0.6725,
+      "step": 5679
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015161368773844403,
+      "loss": 0.741,
+      "step": 5680
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.000151598107049566,
+      "loss": 0.7177,
+      "step": 5681
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015158252465343242,
+      "loss": 0.6734,
+      "step": 5682
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015156694055055877,
+      "loss": 0.7304,
+      "step": 5683
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015155135474146077,
+      "loss": 0.6733,
+      "step": 5684
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001515357672266541,
+      "loss": 0.6673,
+      "step": 5685
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015152017800665451,
+      "loss": 0.6996,
+      "step": 5686
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001515045870819778,
+      "loss": 0.6494,
+      "step": 5687
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015148899445313981,
+      "loss": 0.7323,
+      "step": 5688
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015147340012065653,
+      "loss": 0.6509,
+      "step": 5689
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015145780408504388,
+      "loss": 0.7218,
+      "step": 5690
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015144220634681795,
+      "loss": 0.7129,
+      "step": 5691
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015142660690649478,
+      "loss": 0.7788,
+      "step": 5692
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015141100576459052,
+      "loss": 0.7135,
+      "step": 5693
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001513954029216214,
+      "loss": 0.725,
+      "step": 5694
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001513797983781037,
+      "loss": 0.6612,
+      "step": 5695
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015136419213455365,
+      "loss": 0.717,
+      "step": 5696
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015134858419148768,
+      "loss": 0.6983,
+      "step": 5697
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015133297454942226,
+      "loss": 0.7222,
+      "step": 5698
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001513173632088738,
+      "loss": 0.7416,
+      "step": 5699
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001513017501703589,
+      "loss": 0.7109,
+      "step": 5700
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001512861354343941,
+      "loss": 0.6976,
+      "step": 5701
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001512705190014961,
+      "loss": 0.692,
+      "step": 5702
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001512549008721816,
+      "loss": 0.6511,
+      "step": 5703
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015123928104696734,
+      "loss": 0.683,
+      "step": 5704
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015122365952637017,
+      "loss": 0.6943,
+      "step": 5705
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015120803631090695,
+      "loss": 0.7531,
+      "step": 5706
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015119241140109467,
+      "loss": 0.6709,
+      "step": 5707
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015117678479745022,
+      "loss": 0.7227,
+      "step": 5708
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001511611565004907,
+      "loss": 0.6956,
+      "step": 5709
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015114552651073325,
+      "loss": 0.7029,
+      "step": 5710
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015112989482869498,
+      "loss": 0.6964,
+      "step": 5711
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001511142614548931,
+      "loss": 0.7074,
+      "step": 5712
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001510986263898449,
+      "loss": 0.714,
+      "step": 5713
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001510829896340677,
+      "loss": 0.7128,
+      "step": 5714
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001510673511880789,
+      "loss": 0.7284,
+      "step": 5715
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001510517110523959,
+      "loss": 0.6557,
+      "step": 5716
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001510360692275362,
+      "loss": 0.7105,
+      "step": 5717
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001510204257140174,
+      "loss": 0.661,
+      "step": 5718
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015100478051235705,
+      "loss": 0.6846,
+      "step": 5719
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015098913362307284,
+      "loss": 0.7302,
+      "step": 5720
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001509734850466825,
+      "loss": 0.7429,
+      "step": 5721
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00015095783478370374,
+      "loss": 0.6727,
+      "step": 5722
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015094218283465445,
+      "loss": 0.6994,
+      "step": 5723
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001509265292000525,
+      "loss": 0.6899,
+      "step": 5724
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015091087388041581,
+      "loss": 0.7679,
+      "step": 5725
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015089521687626243,
+      "loss": 0.7286,
+      "step": 5726
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015087955818811037,
+      "loss": 0.6608,
+      "step": 5727
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001508638978164777,
+      "loss": 0.7242,
+      "step": 5728
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001508482357618827,
+      "loss": 0.6788,
+      "step": 5729
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015083257202484348,
+      "loss": 0.6482,
+      "step": 5730
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015081690660587836,
+      "loss": 0.6766,
+      "step": 5731
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015080123950550562,
+      "loss": 0.6688,
+      "step": 5732
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015078557072424373,
+      "loss": 0.7271,
+      "step": 5733
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015076990026261107,
+      "loss": 0.7041,
+      "step": 5734
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001507542281211262,
+      "loss": 0.7208,
+      "step": 5735
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015073855430030756,
+      "loss": 0.6652,
+      "step": 5736
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015072287880067386,
+      "loss": 0.7469,
+      "step": 5737
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015070720162274373,
+      "loss": 0.7166,
+      "step": 5738
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015069152276703593,
+      "loss": 0.7184,
+      "step": 5739
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015067584223406918,
+      "loss": 0.7553,
+      "step": 5740
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001506601600243623,
+      "loss": 0.7287,
+      "step": 5741
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015064447613843423,
+      "loss": 0.6969,
+      "step": 5742
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015062879057680385,
+      "loss": 0.6912,
+      "step": 5743
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015061310333999023,
+      "loss": 0.6803,
+      "step": 5744
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001505974144285124,
+      "loss": 0.6983,
+      "step": 5745
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001505817238428894,
+      "loss": 0.7334,
+      "step": 5746
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001505660315836405,
+      "loss": 0.7337,
+      "step": 5747
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001505503376512848,
+      "loss": 0.7422,
+      "step": 5748
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001505346420463417,
+      "loss": 0.6568,
+      "step": 5749
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015051894476933043,
+      "loss": 0.7088,
+      "step": 5750
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015050324582077047,
+      "loss": 0.6751,
+      "step": 5751
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015048754520118114,
+      "loss": 0.7372,
+      "step": 5752
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015047184291108203,
+      "loss": 0.7518,
+      "step": 5753
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015045613895099265,
+      "loss": 0.7284,
+      "step": 5754
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001504404333214326,
+      "loss": 0.7324,
+      "step": 5755
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001504247260229216,
+      "loss": 0.723,
+      "step": 5756
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015040901705597927,
+      "loss": 0.7773,
+      "step": 5757
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015039330642112545,
+      "loss": 0.627,
+      "step": 5758
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015037759411887995,
+      "loss": 0.6988,
+      "step": 5759
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015036188014976265,
+      "loss": 0.7303,
+      "step": 5760
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015034616451429353,
+      "loss": 0.6967,
+      "step": 5761
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001503304472129925,
+      "loss": 0.7388,
+      "step": 5762
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015031472824637967,
+      "loss": 0.6789,
+      "step": 5763
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015029900761497506,
+      "loss": 0.732,
+      "step": 5764
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015028328531929895,
+      "loss": 0.6603,
+      "step": 5765
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00015026756135987145,
+      "loss": 0.7064,
+      "step": 5766
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015025183573721292,
+      "loss": 0.7027,
+      "step": 5767
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001502361084518436,
+      "loss": 0.7784,
+      "step": 5768
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015022037950428386,
+      "loss": 0.7254,
+      "step": 5769
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001502046488950542,
+      "loss": 0.6812,
+      "step": 5770
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015018891662467508,
+      "loss": 0.6983,
+      "step": 5771
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015017318269366708,
+      "loss": 0.6897,
+      "step": 5772
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001501574471025507,
+      "loss": 0.6675,
+      "step": 5773
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015014170985184667,
+      "loss": 0.6939,
+      "step": 5774
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015012597094207566,
+      "loss": 0.664,
+      "step": 5775
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015011023037375844,
+      "loss": 0.6432,
+      "step": 5776
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001500944881474159,
+      "loss": 0.7147,
+      "step": 5777
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001500787442635688,
+      "loss": 0.7011,
+      "step": 5778
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015006299872273812,
+      "loss": 0.7045,
+      "step": 5779
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015004725152544484,
+      "loss": 0.7526,
+      "step": 5780
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015003150267220997,
+      "loss": 0.7562,
+      "step": 5781
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015001575216355465,
+      "loss": 0.7324,
+      "step": 5782
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00015000000000000001,
+      "loss": 0.6731,
+      "step": 5783
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014998424618206722,
+      "loss": 0.6366,
+      "step": 5784
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014996849071027759,
+      "loss": 0.7265,
+      "step": 5785
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014995273358515236,
+      "loss": 0.7293,
+      "step": 5786
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001499369748072129,
+      "loss": 0.6169,
+      "step": 5787
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014992121437698073,
+      "loss": 0.6889,
+      "step": 5788
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014990545229497722,
+      "loss": 0.6897,
+      "step": 5789
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014988968856172394,
+      "loss": 0.7054,
+      "step": 5790
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014987392317774247,
+      "loss": 0.7071,
+      "step": 5791
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001498581561435544,
+      "loss": 0.7374,
+      "step": 5792
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014984238745968152,
+      "loss": 0.7247,
+      "step": 5793
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014982661712664553,
+      "loss": 0.7245,
+      "step": 5794
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014981084514496823,
+      "loss": 0.7421,
+      "step": 5795
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014979507151517145,
+      "loss": 0.7093,
+      "step": 5796
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014977929623777712,
+      "loss": 0.7565,
+      "step": 5797
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014976351931330722,
+      "loss": 0.6528,
+      "step": 5798
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014974774074228377,
+      "loss": 0.7436,
+      "step": 5799
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001497319605252288,
+      "loss": 0.687,
+      "step": 5800
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014971617866266448,
+      "loss": 0.6368,
+      "step": 5801
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014970039515511304,
+      "loss": 0.7235,
+      "step": 5802
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014968461000309662,
+      "loss": 0.7643,
+      "step": 5803
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014966882320713757,
+      "loss": 0.6908,
+      "step": 5804
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0001496530347677582,
+      "loss": 0.7563,
+      "step": 5805
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014963724468548096,
+      "loss": 0.6949,
+      "step": 5806
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014962145296082828,
+      "loss": 0.7119,
+      "step": 5807
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014960565959432266,
+      "loss": 0.7721,
+      "step": 5808
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00014958986458648668,
+      "loss": 0.6802,
+      "step": 5809
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014957406793784293,
+      "loss": 0.7157,
+      "step": 5810
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014955826964891414,
+      "loss": 0.707,
+      "step": 5811
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014954246972022296,
+      "loss": 0.7236,
+      "step": 5812
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014952666815229223,
+      "loss": 0.7575,
+      "step": 5813
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014951086494564478,
+      "loss": 0.7226,
+      "step": 5814
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014949506010080346,
+      "loss": 0.7179,
+      "step": 5815
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014947925361829127,
+      "loss": 0.7269,
+      "step": 5816
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0001494634454986312,
+      "loss": 0.6439,
+      "step": 5817
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014944763574234623,
+      "loss": 0.7114,
+      "step": 5818
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014943182434995955,
+      "loss": 0.7034,
+      "step": 5819
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014941601132199427,
+      "loss": 0.7273,
+      "step": 5820
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0001494001966589736,
+      "loss": 0.7028,
+      "step": 5821
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014938438036142087,
+      "loss": 0.7548,
+      "step": 5822
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014936856242985937,
+      "loss": 0.6326,
+      "step": 5823
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014935274286481245,
+      "loss": 0.7072,
+      "step": 5824
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014933692166680355,
+      "loss": 0.7242,
+      "step": 5825
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014932109883635618,
+      "loss": 0.7017,
+      "step": 5826
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0001493052743739939,
+      "loss": 0.6424,
+      "step": 5827
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014928944828024023,
+      "loss": 0.7062,
+      "step": 5828
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014927362055561883,
+      "loss": 0.7469,
+      "step": 5829
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0001492577912006535,
+      "loss": 0.7117,
+      "step": 5830
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014924196021586785,
+      "loss": 0.7911,
+      "step": 5831
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014922612760178577,
+      "loss": 0.7231,
+      "step": 5832
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014921029335893113,
+      "loss": 0.6614,
+      "step": 5833
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014919445748782782,
+      "loss": 0.6983,
+      "step": 5834
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014917861998899984,
+      "loss": 0.6791,
+      "step": 5835
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014916278086297117,
+      "loss": 0.6506,
+      "step": 5836
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014914694011026593,
+      "loss": 0.646,
+      "step": 5837
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014913109773140822,
+      "loss": 0.6768,
+      "step": 5838
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014911525372692224,
+      "loss": 0.6995,
+      "step": 5839
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014909940809733222,
+      "loss": 0.6355,
+      "step": 5840
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0001490835608431625,
+      "loss": 0.7389,
+      "step": 5841
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014906771196493735,
+      "loss": 0.6835,
+      "step": 5842
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0001490518614631812,
+      "loss": 0.7012,
+      "step": 5843
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014903600933841853,
+      "loss": 0.7002,
+      "step": 5844
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014902015559117385,
+      "loss": 0.6958,
+      "step": 5845
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014900430022197168,
+      "loss": 0.7232,
+      "step": 5846
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0001489884432313367,
+      "loss": 0.7206,
+      "step": 5847
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014897258461979348,
+      "loss": 0.7019,
+      "step": 5848
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0001489567243878668,
+      "loss": 0.7194,
+      "step": 5849
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014894086253608148,
+      "loss": 0.6698,
+      "step": 5850
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014892499906496228,
+      "loss": 0.7083,
+      "step": 5851
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00014890913397503413,
+      "loss": 0.7616,
+      "step": 5852
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014889326726682197,
+      "loss": 0.6664,
+      "step": 5853
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001488773989408507,
+      "loss": 0.6983,
+      "step": 5854
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014886152899764546,
+      "loss": 0.7577,
+      "step": 5855
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014884565743773136,
+      "loss": 0.6901,
+      "step": 5856
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001488297842616335,
+      "loss": 0.7298,
+      "step": 5857
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014881390946987707,
+      "loss": 0.738,
+      "step": 5858
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014879803306298736,
+      "loss": 0.6938,
+      "step": 5859
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014878215504148968,
+      "loss": 0.7276,
+      "step": 5860
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014876627540590941,
+      "loss": 0.7702,
+      "step": 5861
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014875039415677196,
+      "loss": 0.7334,
+      "step": 5862
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014873451129460277,
+      "loss": 0.6758,
+      "step": 5863
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001487186268199274,
+      "loss": 0.6236,
+      "step": 5864
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014870274073327137,
+      "loss": 0.688,
+      "step": 5865
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001486868530351604,
+      "loss": 0.6837,
+      "step": 5866
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014867096372612012,
+      "loss": 0.7247,
+      "step": 5867
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014865507280667628,
+      "loss": 0.6831,
+      "step": 5868
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014863918027735467,
+      "loss": 0.6854,
+      "step": 5869
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014862328613868115,
+      "loss": 0.7122,
+      "step": 5870
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014860739039118158,
+      "loss": 0.7541,
+      "step": 5871
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014859149303538197,
+      "loss": 0.7668,
+      "step": 5872
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001485755940718083,
+      "loss": 0.6887,
+      "step": 5873
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014855969350098663,
+      "loss": 0.6874,
+      "step": 5874
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.000148543791323443,
+      "loss": 0.7115,
+      "step": 5875
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014852788753970367,
+      "loss": 0.7424,
+      "step": 5876
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014851198215029484,
+      "loss": 0.7182,
+      "step": 5877
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014849607515574276,
+      "loss": 0.7542,
+      "step": 5878
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014848016655657375,
+      "loss": 0.6991,
+      "step": 5879
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014846425635331418,
+      "loss": 0.7424,
+      "step": 5880
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001484483445464905,
+      "loss": 0.7358,
+      "step": 5881
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014843243113662918,
+      "loss": 0.7429,
+      "step": 5882
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001484165161242568,
+      "loss": 0.6949,
+      "step": 5883
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001484005995098999,
+      "loss": 0.74,
+      "step": 5884
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001483846812940851,
+      "loss": 0.8488,
+      "step": 5885
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014836876147733918,
+      "loss": 0.7137,
+      "step": 5886
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014835284006018882,
+      "loss": 0.6787,
+      "step": 5887
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001483369170431608,
+      "loss": 0.6464,
+      "step": 5888
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014832099242678205,
+      "loss": 0.6492,
+      "step": 5889
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014830506621157946,
+      "loss": 0.6915,
+      "step": 5890
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014828913839807993,
+      "loss": 0.7693,
+      "step": 5891
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0001482732089868105,
+      "loss": 0.7087,
+      "step": 5892
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014825727797829826,
+      "loss": 0.7293,
+      "step": 5893
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014824134537307033,
+      "loss": 0.8153,
+      "step": 5894
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00014822541117165384,
+      "loss": 0.7319,
+      "step": 5895
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014820947537457608,
+      "loss": 0.7222,
+      "step": 5896
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014819353798236427,
+      "loss": 0.7504,
+      "step": 5897
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014817759899554572,
+      "loss": 0.6521,
+      "step": 5898
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014816165841464787,
+      "loss": 0.7267,
+      "step": 5899
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014814571624019814,
+      "loss": 0.7216,
+      "step": 5900
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014812977247272402,
+      "loss": 0.7,
+      "step": 5901
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014811382711275303,
+      "loss": 0.735,
+      "step": 5902
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014809788016081275,
+      "loss": 0.7547,
+      "step": 5903
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001480819316174309,
+      "loss": 0.811,
+      "step": 5904
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001480659814831351,
+      "loss": 0.76,
+      "step": 5905
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014805002975845313,
+      "loss": 0.6951,
+      "step": 5906
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001480340764439128,
+      "loss": 0.7336,
+      "step": 5907
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014801812154004192,
+      "loss": 0.6764,
+      "step": 5908
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014800216504736848,
+      "loss": 0.7144,
+      "step": 5909
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014798620696642042,
+      "loss": 0.7043,
+      "step": 5910
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014797024729772568,
+      "loss": 0.7506,
+      "step": 5911
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014795428604181242,
+      "loss": 0.6277,
+      "step": 5912
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014793832319920868,
+      "loss": 0.7065,
+      "step": 5913
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001479223587704427,
+      "loss": 0.7686,
+      "step": 5914
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014790639275604267,
+      "loss": 0.7066,
+      "step": 5915
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014789042515653687,
+      "loss": 0.7728,
+      "step": 5916
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014787445597245363,
+      "loss": 0.7263,
+      "step": 5917
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014785848520432133,
+      "loss": 0.6765,
+      "step": 5918
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001478425128526684,
+      "loss": 0.6948,
+      "step": 5919
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014782653891802334,
+      "loss": 0.6908,
+      "step": 5920
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001478105634009147,
+      "loss": 0.7445,
+      "step": 5921
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.000147794586301871,
+      "loss": 0.7468,
+      "step": 5922
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014777860762142097,
+      "loss": 0.7001,
+      "step": 5923
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014776262736009325,
+      "loss": 0.7724,
+      "step": 5924
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001477466455184166,
+      "loss": 0.7378,
+      "step": 5925
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014773066209691988,
+      "loss": 0.694,
+      "step": 5926
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014771467709613184,
+      "loss": 0.7426,
+      "step": 5927
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014769869051658148,
+      "loss": 0.7348,
+      "step": 5928
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014768270235879767,
+      "loss": 0.6756,
+      "step": 5929
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014766671262330948,
+      "loss": 0.7289,
+      "step": 5930
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014765072131064593,
+      "loss": 0.717,
+      "step": 5931
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001476347284213362,
+      "loss": 0.7583,
+      "step": 5932
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014761873395590936,
+      "loss": 0.7393,
+      "step": 5933
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014760273791489473,
+      "loss": 0.7225,
+      "step": 5934
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014758674029882152,
+      "loss": 0.737,
+      "step": 5935
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014757074110821904,
+      "loss": 0.6843,
+      "step": 5936
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00014755474034361672,
+      "loss": 0.7388,
+      "step": 5937
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001475387380055439,
+      "loss": 0.781,
+      "step": 5938
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014752273409453015,
+      "loss": 0.6595,
+      "step": 5939
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014750672861110496,
+      "loss": 0.6743,
+      "step": 5940
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014749072155579787,
+      "loss": 0.6835,
+      "step": 5941
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014747471292913863,
+      "loss": 0.6924,
+      "step": 5942
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014745870273165677,
+      "loss": 0.6698,
+      "step": 5943
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014744269096388218,
+      "loss": 0.7044,
+      "step": 5944
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014742667762634456,
+      "loss": 0.6533,
+      "step": 5945
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014741066271957377,
+      "loss": 0.7347,
+      "step": 5946
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014739464624409976,
+      "loss": 0.6451,
+      "step": 5947
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0001473786282004524,
+      "loss": 0.6624,
+      "step": 5948
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014736260858916165,
+      "loss": 0.6967,
+      "step": 5949
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014734658741075767,
+      "loss": 0.7512,
+      "step": 5950
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0001473305646657705,
+      "loss": 0.7577,
+      "step": 5951
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014731454035473034,
+      "loss": 0.6889,
+      "step": 5952
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014729851447816737,
+      "loss": 0.6994,
+      "step": 5953
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014728248703661182,
+      "loss": 0.6983,
+      "step": 5954
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.000147266458030594,
+      "loss": 0.6747,
+      "step": 5955
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014725042746064433,
+      "loss": 0.6663,
+      "step": 5956
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014723439532729317,
+      "loss": 0.72,
+      "step": 5957
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014721836163107102,
+      "loss": 0.7151,
+      "step": 5958
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014720232637250834,
+      "loss": 0.7233,
+      "step": 5959
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014718628955213574,
+      "loss": 0.6802,
+      "step": 5960
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014717025117048381,
+      "loss": 0.7075,
+      "step": 5961
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0001471542112280833,
+      "loss": 0.7419,
+      "step": 5962
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014713816972546483,
+      "loss": 0.6998,
+      "step": 5963
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014712212666315923,
+      "loss": 0.6986,
+      "step": 5964
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014710608204169734,
+      "loss": 0.6324,
+      "step": 5965
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014709003586160994,
+      "loss": 0.7197,
+      "step": 5966
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014707398812342808,
+      "loss": 0.6438,
+      "step": 5967
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0001470579388276827,
+      "loss": 0.7146,
+      "step": 5968
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014704188797490483,
+      "loss": 0.7217,
+      "step": 5969
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014702583556562554,
+      "loss": 0.7562,
+      "step": 5970
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014700978160037595,
+      "loss": 0.7289,
+      "step": 5971
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014699372607968725,
+      "loss": 0.7944,
+      "step": 5972
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014697766900409074,
+      "loss": 0.7116,
+      "step": 5973
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014696161037411768,
+      "loss": 0.6528,
+      "step": 5974
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014694555019029938,
+      "loss": 0.6887,
+      "step": 5975
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014692948845316722,
+      "loss": 0.6619,
+      "step": 5976
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014691342516325268,
+      "loss": 0.7516,
+      "step": 5977
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014689736032108728,
+      "loss": 0.6155,
+      "step": 5978
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014688129392720253,
+      "loss": 0.7902,
+      "step": 5979
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014686522598213002,
+      "loss": 0.7143,
+      "step": 5980
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00014684915648640142,
+      "loss": 0.6633,
+      "step": 5981
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0001468330854405484,
+      "loss": 0.6982,
+      "step": 5982
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014681701284510274,
+      "loss": 0.6571,
+      "step": 5983
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014680093870059625,
+      "loss": 0.6799,
+      "step": 5984
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0001467848630075608,
+      "loss": 0.7502,
+      "step": 5985
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014676878576652823,
+      "loss": 0.6953,
+      "step": 5986
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014675270697803053,
+      "loss": 0.6634,
+      "step": 5987
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014673662664259975,
+      "loss": 0.7359,
+      "step": 5988
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014672054476076783,
+      "loss": 0.8505,
+      "step": 5989
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014670446133306707,
+      "loss": 0.7246,
+      "step": 5990
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014668837636002945,
+      "loss": 0.75,
+      "step": 5991
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0001466722898421873,
+      "loss": 0.7024,
+      "step": 5992
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014665620178007277,
+      "loss": 0.6771,
+      "step": 5993
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014664011217421828,
+      "loss": 0.7571,
+      "step": 5994
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014662402102515618,
+      "loss": 0.7231,
+      "step": 5995
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014660792833341883,
+      "loss": 0.6522,
+      "step": 5996
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014659183409953877,
+      "loss": 0.6647,
+      "step": 5997
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014657573832404846,
+      "loss": 0.7596,
+      "step": 5998
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014655964100748048,
+      "loss": 0.7084,
+      "step": 5999
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0001465435421503675,
+      "loss": 0.7394,
+      "step": 6000
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014652744175324217,
+      "loss": 0.6953,
+      "step": 6001
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014651133981663714,
+      "loss": 0.7161,
+      "step": 6002
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014649523634108525,
+      "loss": 0.724,
+      "step": 6003
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014647913132711935,
+      "loss": 0.7388,
+      "step": 6004
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014646302477527227,
+      "loss": 0.7458,
+      "step": 6005
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014644691668607695,
+      "loss": 0.7913,
+      "step": 6006
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014643080706006637,
+      "loss": 0.7001,
+      "step": 6007
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014641469589777355,
+      "loss": 0.7269,
+      "step": 6008
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0001463985831997316,
+      "loss": 0.7354,
+      "step": 6009
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0001463824689664736,
+      "loss": 0.6775,
+      "step": 6010
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014636635319853275,
+      "loss": 0.6977,
+      "step": 6011
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014635023589644228,
+      "loss": 0.7233,
+      "step": 6012
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0001463341170607355,
+      "loss": 0.6738,
+      "step": 6013
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0001463179966919457,
+      "loss": 0.8296,
+      "step": 6014
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014630187479060632,
+      "loss": 0.6844,
+      "step": 6015
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014628575135725074,
+      "loss": 0.6772,
+      "step": 6016
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014626962639241247,
+      "loss": 0.7277,
+      "step": 6017
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014625349989662501,
+      "loss": 0.7232,
+      "step": 6018
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.000146237371870422,
+      "loss": 0.7476,
+      "step": 6019
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014622124231433706,
+      "loss": 0.6425,
+      "step": 6020
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014620511122890388,
+      "loss": 0.6204,
+      "step": 6021
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014618897861465617,
+      "loss": 0.6774,
+      "step": 6022
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014617284447212775,
+      "loss": 0.7085,
+      "step": 6023
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014615670880185243,
+      "loss": 0.6717,
+      "step": 6024
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00014614057160436414,
+      "loss": 0.7391,
+      "step": 6025
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0001461244328801968,
+      "loss": 0.7701,
+      "step": 6026
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0001461082926298844,
+      "loss": 0.714,
+      "step": 6027
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014609215085396095,
+      "loss": 0.7074,
+      "step": 6028
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014607600755296062,
+      "loss": 0.7072,
+      "step": 6029
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014605986272741748,
+      "loss": 0.7122,
+      "step": 6030
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014604371637786577,
+      "loss": 0.7186,
+      "step": 6031
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014602756850483968,
+      "loss": 0.7003,
+      "step": 6032
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014601141910887354,
+      "loss": 0.7549,
+      "step": 6033
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0001459952681905017,
+      "loss": 0.7385,
+      "step": 6034
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014597911575025855,
+      "loss": 0.7601,
+      "step": 6035
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014596296178867854,
+      "loss": 0.7899,
+      "step": 6036
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014594680630629613,
+      "loss": 0.6542,
+      "step": 6037
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014593064930364593,
+      "loss": 0.7587,
+      "step": 6038
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014591449078126244,
+      "loss": 0.7573,
+      "step": 6039
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0001458983307396804,
+      "loss": 0.6571,
+      "step": 6040
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014588216917943447,
+      "loss": 0.6958,
+      "step": 6041
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014586600610105938,
+      "loss": 0.8214,
+      "step": 6042
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014584984150508998,
+      "loss": 0.663,
+      "step": 6043
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014583367539206102,
+      "loss": 0.6907,
+      "step": 6044
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014581750776250749,
+      "loss": 0.6696,
+      "step": 6045
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0001458013386169643,
+      "loss": 0.7819,
+      "step": 6046
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014578516795596644,
+      "loss": 0.7336,
+      "step": 6047
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.000145768995780049,
+      "loss": 0.7442,
+      "step": 6048
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014575282208974702,
+      "loss": 0.6542,
+      "step": 6049
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014573664688559567,
+      "loss": 0.7158,
+      "step": 6050
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014572047016813019,
+      "loss": 0.6901,
+      "step": 6051
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014570429193788576,
+      "loss": 0.775,
+      "step": 6052
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0001456881121953977,
+      "loss": 0.7249,
+      "step": 6053
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014567193094120143,
+      "loss": 0.66,
+      "step": 6054
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014565574817583223,
+      "loss": 0.6439,
+      "step": 6055
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0001456395638998256,
+      "loss": 0.72,
+      "step": 6056
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014562337811371707,
+      "loss": 0.6665,
+      "step": 6057
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014560719081804216,
+      "loss": 0.7122,
+      "step": 6058
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014559100201333647,
+      "loss": 0.6459,
+      "step": 6059
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014557481170013564,
+      "loss": 0.6592,
+      "step": 6060
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014555861987897536,
+      "loss": 0.69,
+      "step": 6061
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014554242655039139,
+      "loss": 0.7288,
+      "step": 6062
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014552623171491954,
+      "loss": 0.8172,
+      "step": 6063
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014551003537309568,
+      "loss": 0.6955,
+      "step": 6064
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014549383752545566,
+      "loss": 0.7604,
+      "step": 6065
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014547763817253544,
+      "loss": 0.6631,
+      "step": 6066
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.000145461437314871,
+      "loss": 0.7376,
+      "step": 6067
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00014544523495299842,
+      "loss": 0.7163,
+      "step": 6068
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014542903108745377,
+      "loss": 0.7116,
+      "step": 6069
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014541282571877326,
+      "loss": 0.7457,
+      "step": 6070
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.000145396618847493,
+      "loss": 0.7392,
+      "step": 6071
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014538041047414927,
+      "loss": 0.6931,
+      "step": 6072
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014536420059927838,
+      "loss": 0.66,
+      "step": 6073
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014534798922341666,
+      "loss": 0.7095,
+      "step": 6074
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014533177634710052,
+      "loss": 0.7376,
+      "step": 6075
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014531556197086638,
+      "loss": 0.767,
+      "step": 6076
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014529934609525076,
+      "loss": 0.7177,
+      "step": 6077
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014528312872079017,
+      "loss": 0.7036,
+      "step": 6078
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001452669098480212,
+      "loss": 0.6955,
+      "step": 6079
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014525068947748057,
+      "loss": 0.6893,
+      "step": 6080
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014523446760970492,
+      "loss": 0.7379,
+      "step": 6081
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014521824424523096,
+      "loss": 0.7407,
+      "step": 6082
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014520201938459553,
+      "loss": 0.7072,
+      "step": 6083
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014518579302833542,
+      "loss": 0.6929,
+      "step": 6084
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001451695651769876,
+      "loss": 0.6782,
+      "step": 6085
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014515333583108896,
+      "loss": 0.7194,
+      "step": 6086
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014513710499117647,
+      "loss": 0.7415,
+      "step": 6087
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001451208726577872,
+      "loss": 0.7028,
+      "step": 6088
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014510463883145826,
+      "loss": 0.6425,
+      "step": 6089
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001450884035127267,
+      "loss": 0.7448,
+      "step": 6090
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014507216670212985,
+      "loss": 0.7297,
+      "step": 6091
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014505592840020478,
+      "loss": 0.6732,
+      "step": 6092
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014503968860748892,
+      "loss": 0.7028,
+      "step": 6093
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001450234473245195,
+      "loss": 0.6965,
+      "step": 6094
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.000145007204551834,
+      "loss": 0.7266,
+      "step": 6095
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014499096028996978,
+      "loss": 0.7451,
+      "step": 6096
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014497471453946432,
+      "loss": 0.6882,
+      "step": 6097
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014495846730085517,
+      "loss": 0.6677,
+      "step": 6098
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014494221857467996,
+      "loss": 0.6013,
+      "step": 6099
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014492596836147628,
+      "loss": 0.6797,
+      "step": 6100
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014490971666178176,
+      "loss": 0.6986,
+      "step": 6101
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014489346347613423,
+      "loss": 0.6555,
+      "step": 6102
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001448772088050714,
+      "loss": 0.7071,
+      "step": 6103
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014486095264913112,
+      "loss": 0.6817,
+      "step": 6104
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014484469500885126,
+      "loss": 0.7136,
+      "step": 6105
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014482843588476974,
+      "loss": 0.6932,
+      "step": 6106
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014481217527742457,
+      "loss": 0.6906,
+      "step": 6107
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001447959131873537,
+      "loss": 0.7784,
+      "step": 6108
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014477964961509527,
+      "loss": 0.7012,
+      "step": 6109
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00014476338456118737,
+      "loss": 0.6845,
+      "step": 6110
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0001447471180261682,
+      "loss": 0.7136,
+      "step": 6111
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014473085001057596,
+      "loss": 0.7504,
+      "step": 6112
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014471458051494893,
+      "loss": 0.6342,
+      "step": 6113
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001446983095398254,
+      "loss": 0.6795,
+      "step": 6114
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014468203708574375,
+      "loss": 0.661,
+      "step": 6115
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001446657631532424,
+      "loss": 0.7034,
+      "step": 6116
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014464948774285983,
+      "loss": 0.622,
+      "step": 6117
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014463321085513454,
+      "loss": 0.747,
+      "step": 6118
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014461693249060508,
+      "loss": 0.6757,
+      "step": 6119
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014460065264981008,
+      "loss": 0.7277,
+      "step": 6120
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001445843713332882,
+      "loss": 0.6845,
+      "step": 6121
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014456808854157815,
+      "loss": 0.6985,
+      "step": 6122
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014455180427521867,
+      "loss": 0.7303,
+      "step": 6123
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014453551853474856,
+      "loss": 0.7602,
+      "step": 6124
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001445192313207067,
+      "loss": 0.6491,
+      "step": 6125
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014450294263363203,
+      "loss": 0.7106,
+      "step": 6126
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001444866524740634,
+      "loss": 0.7445,
+      "step": 6127
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001444703608425399,
+      "loss": 0.7221,
+      "step": 6128
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014445406773960055,
+      "loss": 0.6625,
+      "step": 6129
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014443777316578446,
+      "loss": 0.7524,
+      "step": 6130
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014442147712163077,
+      "loss": 0.7426,
+      "step": 6131
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014440517960767868,
+      "loss": 0.7427,
+      "step": 6132
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001443888806244674,
+      "loss": 0.741,
+      "step": 6133
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014437258017253627,
+      "loss": 0.7162,
+      "step": 6134
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014435627825242463,
+      "loss": 0.7273,
+      "step": 6135
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014433997486467185,
+      "loss": 0.7017,
+      "step": 6136
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014432367000981734,
+      "loss": 0.646,
+      "step": 6137
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014430736368840067,
+      "loss": 0.7011,
+      "step": 6138
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014429105590096133,
+      "loss": 0.6885,
+      "step": 6139
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001442747466480389,
+      "loss": 0.72,
+      "step": 6140
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014425843593017298,
+      "loss": 0.6769,
+      "step": 6141
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014424212374790333,
+      "loss": 0.7561,
+      "step": 6142
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014422581010176962,
+      "loss": 0.6973,
+      "step": 6143
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014420949499231172,
+      "loss": 0.733,
+      "step": 6144
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001441931784200693,
+      "loss": 0.714,
+      "step": 6145
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014417686038558234,
+      "loss": 0.7211,
+      "step": 6146
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014416054088939076,
+      "loss": 0.7744,
+      "step": 6147
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014414421993203452,
+      "loss": 0.6788,
+      "step": 6148
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014412789751405363,
+      "loss": 0.6956,
+      "step": 6149
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014411157363598823,
+      "loss": 0.6502,
+      "step": 6150
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001440952482983783,
+      "loss": 0.6732,
+      "step": 6151
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0001440789215017641,
+      "loss": 0.7097,
+      "step": 6152
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014406259324668583,
+      "loss": 0.6668,
+      "step": 6153
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00014404626353368375,
+      "loss": 0.7255,
+      "step": 6154
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001440299323632982,
+      "loss": 0.6592,
+      "step": 6155
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014401359973606947,
+      "loss": 0.6359,
+      "step": 6156
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014399726565253798,
+      "loss": 0.6552,
+      "step": 6157
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014398093011324425,
+      "loss": 0.7131,
+      "step": 6158
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014396459311872875,
+      "loss": 0.6965,
+      "step": 6159
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014394825466953202,
+      "loss": 0.7241,
+      "step": 6160
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014393191476619465,
+      "loss": 0.7624,
+      "step": 6161
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001439155734092573,
+      "loss": 0.6798,
+      "step": 6162
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014389923059926062,
+      "loss": 0.7129,
+      "step": 6163
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014388288633674544,
+      "loss": 0.7064,
+      "step": 6164
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014386654062225257,
+      "loss": 0.7995,
+      "step": 6165
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014385019345632274,
+      "loss": 0.7205,
+      "step": 6166
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014383384483949687,
+      "loss": 0.7173,
+      "step": 6167
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014381749477231591,
+      "loss": 0.6924,
+      "step": 6168
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014380114325532087,
+      "loss": 0.7491,
+      "step": 6169
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014378479028905278,
+      "loss": 0.7266,
+      "step": 6170
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001437684358740527,
+      "loss": 0.7027,
+      "step": 6171
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014375208001086173,
+      "loss": 0.7044,
+      "step": 6172
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001437357227000211,
+      "loss": 0.7762,
+      "step": 6173
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014371936394207198,
+      "loss": 0.7119,
+      "step": 6174
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014370300373755566,
+      "loss": 0.6717,
+      "step": 6175
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014368664208701352,
+      "loss": 0.6833,
+      "step": 6176
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014367027899098687,
+      "loss": 0.6932,
+      "step": 6177
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001436539144500171,
+      "loss": 0.7082,
+      "step": 6178
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001436375484646457,
+      "loss": 0.7279,
+      "step": 6179
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014362118103541417,
+      "loss": 0.6943,
+      "step": 6180
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014360481216286414,
+      "loss": 0.7114,
+      "step": 6181
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014358844184753712,
+      "loss": 0.6439,
+      "step": 6182
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014357207008997483,
+      "loss": 0.7425,
+      "step": 6183
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001435556968907189,
+      "loss": 0.7103,
+      "step": 6184
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014353932225031113,
+      "loss": 0.7295,
+      "step": 6185
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014352294616929334,
+      "loss": 0.6648,
+      "step": 6186
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014350656864820733,
+      "loss": 0.7833,
+      "step": 6187
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014349018968759503,
+      "loss": 0.6516,
+      "step": 6188
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014347380928799832,
+      "loss": 0.6209,
+      "step": 6189
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014345742744995924,
+      "loss": 0.6938,
+      "step": 6190
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014344104417401979,
+      "loss": 0.6892,
+      "step": 6191
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001434246594607221,
+      "loss": 0.6993,
+      "step": 6192
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014340827331060826,
+      "loss": 0.7302,
+      "step": 6193
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014339188572422046,
+      "loss": 0.7229,
+      "step": 6194
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014337549670210092,
+      "loss": 0.7281,
+      "step": 6195
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014335910624479193,
+      "loss": 0.6999,
+      "step": 6196
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0001433427143528358,
+      "loss": 0.6595,
+      "step": 6197
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00014332632102677487,
+      "loss": 0.7443,
+      "step": 6198
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001433099262671516,
+      "loss": 0.7256,
+      "step": 6199
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014329353007450844,
+      "loss": 0.6759,
+      "step": 6200
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001432771324493879,
+      "loss": 0.7491,
+      "step": 6201
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001432607333923325,
+      "loss": 0.6639,
+      "step": 6202
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001432443329038849,
+      "loss": 0.6831,
+      "step": 6203
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014322793098458774,
+      "loss": 0.6955,
+      "step": 6204
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001432115276349837,
+      "loss": 0.724,
+      "step": 6205
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001431951228556155,
+      "loss": 0.6787,
+      "step": 6206
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.000143178716647026,
+      "loss": 0.7161,
+      "step": 6207
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.000143162309009758,
+      "loss": 0.781,
+      "step": 6208
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001431458999443544,
+      "loss": 0.7076,
+      "step": 6209
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014312948945135814,
+      "loss": 0.6992,
+      "step": 6210
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001431130775313122,
+      "loss": 0.7121,
+      "step": 6211
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014309666418475963,
+      "loss": 0.6688,
+      "step": 6212
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014308024941224345,
+      "loss": 0.6799,
+      "step": 6213
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014306383321430682,
+      "loss": 0.6228,
+      "step": 6214
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014304741559149293,
+      "loss": 0.702,
+      "step": 6215
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014303099654434497,
+      "loss": 0.7666,
+      "step": 6216
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014301457607340625,
+      "loss": 0.7167,
+      "step": 6217
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014299815417922,
+      "loss": 0.6634,
+      "step": 6218
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001429817308623297,
+      "loss": 0.715,
+      "step": 6219
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014296530612327863,
+      "loss": 0.7199,
+      "step": 6220
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014294887996261032,
+      "loss": 0.6805,
+      "step": 6221
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014293245238086827,
+      "loss": 0.7034,
+      "step": 6222
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.000142916023378596,
+      "loss": 0.6886,
+      "step": 6223
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014289959295633712,
+      "loss": 0.7444,
+      "step": 6224
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014288316111463529,
+      "loss": 0.6556,
+      "step": 6225
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001428667278540342,
+      "loss": 0.7657,
+      "step": 6226
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014285029317507753,
+      "loss": 0.6993,
+      "step": 6227
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014283385707830913,
+      "loss": 0.6941,
+      "step": 6228
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014281741956427278,
+      "loss": 0.7725,
+      "step": 6229
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014280098063351238,
+      "loss": 0.757,
+      "step": 6230
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014278454028657188,
+      "loss": 0.6956,
+      "step": 6231
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001427680985239952,
+      "loss": 0.6555,
+      "step": 6232
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014275165534632638,
+      "loss": 0.6513,
+      "step": 6233
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014273521075410948,
+      "loss": 0.7014,
+      "step": 6234
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014271876474788864,
+      "loss": 0.6531,
+      "step": 6235
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.000142702317328208,
+      "loss": 0.6644,
+      "step": 6236
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014268586849561175,
+      "loss": 0.6019,
+      "step": 6237
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014266941825064416,
+      "loss": 0.7851,
+      "step": 6238
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014265296659384956,
+      "loss": 0.7101,
+      "step": 6239
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014263651352577222,
+      "loss": 0.6996,
+      "step": 6240
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00014262005904695657,
+      "loss": 0.7245,
+      "step": 6241
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014260360315794705,
+      "loss": 0.7087,
+      "step": 6242
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014258714585928814,
+      "loss": 0.6688,
+      "step": 6243
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001425706871515244,
+      "loss": 0.767,
+      "step": 6244
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014255422703520035,
+      "loss": 0.6656,
+      "step": 6245
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014253776551086068,
+      "loss": 0.7173,
+      "step": 6246
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014252130257905,
+      "loss": 0.7227,
+      "step": 6247
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014250483824031307,
+      "loss": 0.7543,
+      "step": 6248
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014248837249519468,
+      "loss": 0.6794,
+      "step": 6249
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014247190534423954,
+      "loss": 0.7207,
+      "step": 6250
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014245543678799265,
+      "loss": 0.7171,
+      "step": 6251
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014243896682699877,
+      "loss": 0.6406,
+      "step": 6252
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014242249546180297,
+      "loss": 0.6467,
+      "step": 6253
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014240602269295017,
+      "loss": 0.6746,
+      "step": 6254
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014238954852098545,
+      "loss": 0.7975,
+      "step": 6255
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001423730729464539,
+      "loss": 0.6205,
+      "step": 6256
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014235659596990063,
+      "loss": 0.7418,
+      "step": 6257
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014234011759187083,
+      "loss": 0.7358,
+      "step": 6258
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014232363781290975,
+      "loss": 0.6773,
+      "step": 6259
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014230715663356264,
+      "loss": 0.6946,
+      "step": 6260
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001422906740543749,
+      "loss": 0.6958,
+      "step": 6261
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014227419007589175,
+      "loss": 0.7888,
+      "step": 6262
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014225770469865874,
+      "loss": 0.6772,
+      "step": 6263
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014224121792322125,
+      "loss": 0.65,
+      "step": 6264
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014222472975012482,
+      "loss": 0.7172,
+      "step": 6265
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014220824017991505,
+      "loss": 0.6948,
+      "step": 6266
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014219174921313745,
+      "loss": 0.7707,
+      "step": 6267
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001421752568503377,
+      "loss": 0.7444,
+      "step": 6268
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014215876309206152,
+      "loss": 0.691,
+      "step": 6269
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001421422679388546,
+      "loss": 0.6969,
+      "step": 6270
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001421257713912628,
+      "loss": 0.7209,
+      "step": 6271
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001421092734498319,
+      "loss": 0.6472,
+      "step": 6272
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014209277411510773,
+      "loss": 0.703,
+      "step": 6273
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001420762733876363,
+      "loss": 0.6845,
+      "step": 6274
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001420597712679635,
+      "loss": 0.7607,
+      "step": 6275
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014204326775663542,
+      "loss": 0.7943,
+      "step": 6276
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014202676285419812,
+      "loss": 0.6493,
+      "step": 6277
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0001420102565611976,
+      "loss": 0.6645,
+      "step": 6278
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014199374887818012,
+      "loss": 0.6692,
+      "step": 6279
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014197723980569185,
+      "loss": 0.734,
+      "step": 6280
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000141960729344279,
+      "loss": 0.6963,
+      "step": 6281
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014194421749448796,
+      "loss": 0.6874,
+      "step": 6282
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014192770425686498,
+      "loss": 0.666,
+      "step": 6283
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00014191118963195642,
+      "loss": 0.7274,
+      "step": 6284
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014189467362030875,
+      "loss": 0.6926,
+      "step": 6285
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014187815622246846,
+      "loss": 0.6823,
+      "step": 6286
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014186163743898207,
+      "loss": 0.6356,
+      "step": 6287
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014184511727039612,
+      "loss": 0.7102,
+      "step": 6288
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014182859571725725,
+      "loss": 0.6904,
+      "step": 6289
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014181207278011207,
+      "loss": 0.6699,
+      "step": 6290
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0001417955484595073,
+      "loss": 0.7055,
+      "step": 6291
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014177902275598975,
+      "loss": 0.7343,
+      "step": 6292
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014176249567010614,
+      "loss": 0.7291,
+      "step": 6293
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014174596720240335,
+      "loss": 0.6885,
+      "step": 6294
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014172943735342826,
+      "loss": 0.7321,
+      "step": 6295
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0001417129061237278,
+      "loss": 0.7892,
+      "step": 6296
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014169637351384897,
+      "loss": 0.7261,
+      "step": 6297
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014167983952433874,
+      "loss": 0.6894,
+      "step": 6298
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014166330415574423,
+      "loss": 0.757,
+      "step": 6299
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0001416467674086125,
+      "loss": 0.6634,
+      "step": 6300
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014163022928349078,
+      "loss": 0.659,
+      "step": 6301
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014161368978092624,
+      "loss": 0.7438,
+      "step": 6302
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014159714890146614,
+      "loss": 0.6361,
+      "step": 6303
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014158060664565776,
+      "loss": 0.7331,
+      "step": 6304
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014156406301404846,
+      "loss": 0.7088,
+      "step": 6305
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014154751800718563,
+      "loss": 0.6894,
+      "step": 6306
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014153097162561667,
+      "loss": 0.738,
+      "step": 6307
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0001415144238698891,
+      "loss": 0.7268,
+      "step": 6308
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014149787474055045,
+      "loss": 0.6966,
+      "step": 6309
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014148132423814826,
+      "loss": 0.7005,
+      "step": 6310
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014146477236323015,
+      "loss": 0.712,
+      "step": 6311
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0001414482191163438,
+      "loss": 0.7486,
+      "step": 6312
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0001414316644980369,
+      "loss": 0.7154,
+      "step": 6313
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014141510850885721,
+      "loss": 0.6991,
+      "step": 6314
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014139855114935252,
+      "loss": 0.6349,
+      "step": 6315
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014138199242007068,
+      "loss": 0.6589,
+      "step": 6316
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014136543232155959,
+      "loss": 0.7141,
+      "step": 6317
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014134887085436715,
+      "loss": 0.7827,
+      "step": 6318
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014133230801904135,
+      "loss": 0.7045,
+      "step": 6319
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014131574381613026,
+      "loss": 0.6837,
+      "step": 6320
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0001412991782461819,
+      "loss": 0.6418,
+      "step": 6321
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014128261130974437,
+      "loss": 0.7203,
+      "step": 6322
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014126604300736584,
+      "loss": 0.6475,
+      "step": 6323
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014124947333959457,
+      "loss": 0.6641,
+      "step": 6324
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014123290230697876,
+      "loss": 0.7823,
+      "step": 6325
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0001412163299100667,
+      "loss": 0.7543,
+      "step": 6326
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00014119975614940673,
+      "loss": 0.6959,
+      "step": 6327
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014118318102554727,
+      "loss": 0.7386,
+      "step": 6328
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014116660453903677,
+      "loss": 0.6719,
+      "step": 6329
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014115002669042362,
+      "loss": 0.663,
+      "step": 6330
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001411334474802564,
+      "loss": 0.7249,
+      "step": 6331
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014111686690908366,
+      "loss": 0.7121,
+      "step": 6332
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.000141100284977454,
+      "loss": 0.6275,
+      "step": 6333
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001410837016859161,
+      "loss": 0.6827,
+      "step": 6334
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014106711703501865,
+      "loss": 0.6529,
+      "step": 6335
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014105053102531038,
+      "loss": 0.625,
+      "step": 6336
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001410339436573401,
+      "loss": 0.7342,
+      "step": 6337
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014101735493165667,
+      "loss": 0.7584,
+      "step": 6338
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014100076484880892,
+      "loss": 0.6781,
+      "step": 6339
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014098417340934578,
+      "loss": 0.687,
+      "step": 6340
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014096758061381628,
+      "loss": 0.633,
+      "step": 6341
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014095098646276934,
+      "loss": 0.6642,
+      "step": 6342
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001409343909567541,
+      "loss": 0.7228,
+      "step": 6343
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014091779409631963,
+      "loss": 0.6827,
+      "step": 6344
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001409011958820151,
+      "loss": 0.7583,
+      "step": 6345
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014088459631438971,
+      "loss": 0.6696,
+      "step": 6346
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014086799539399266,
+      "loss": 0.6697,
+      "step": 6347
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001408513931213732,
+      "loss": 0.7302,
+      "step": 6348
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014083478949708078,
+      "loss": 0.6334,
+      "step": 6349
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014081818452166468,
+      "loss": 0.7318,
+      "step": 6350
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014080157819567433,
+      "loss": 0.6693,
+      "step": 6351
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014078497051965927,
+      "loss": 0.6558,
+      "step": 6352
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014076836149416887,
+      "loss": 0.7387,
+      "step": 6353
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014075175111975282,
+      "loss": 0.7243,
+      "step": 6354
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014073513939696062,
+      "loss": 0.701,
+      "step": 6355
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014071852632634196,
+      "loss": 0.6341,
+      "step": 6356
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001407019119084465,
+      "loss": 0.733,
+      "step": 6357
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.000140685296143824,
+      "loss": 0.7024,
+      "step": 6358
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001406686790330242,
+      "loss": 0.7386,
+      "step": 6359
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014065206057659691,
+      "loss": 0.7632,
+      "step": 6360
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014063544077509206,
+      "loss": 0.6716,
+      "step": 6361
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014061881962905954,
+      "loss": 0.6655,
+      "step": 6362
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014060219713904926,
+      "loss": 0.7099,
+      "step": 6363
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014058557330561122,
+      "loss": 0.7505,
+      "step": 6364
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001405689481292955,
+      "loss": 0.6202,
+      "step": 6365
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014055232161065215,
+      "loss": 0.7058,
+      "step": 6366
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014053569375023134,
+      "loss": 0.6213,
+      "step": 6367
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014051906454858325,
+      "loss": 0.7275,
+      "step": 6368
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014050243400625805,
+      "loss": 0.7527,
+      "step": 6369
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00014048580212380602,
+      "loss": 0.6691,
+      "step": 6370
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001404691689017775,
+      "loss": 0.7226,
+      "step": 6371
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001404525343407228,
+      "loss": 0.7103,
+      "step": 6372
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014043589844119235,
+      "loss": 0.7006,
+      "step": 6373
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001404192612037366,
+      "loss": 0.6689,
+      "step": 6374
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014040262262890598,
+      "loss": 0.7079,
+      "step": 6375
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014038598271725103,
+      "loss": 0.7385,
+      "step": 6376
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014036934146932238,
+      "loss": 0.7059,
+      "step": 6377
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014035269888567066,
+      "loss": 0.7798,
+      "step": 6378
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014033605496684644,
+      "loss": 0.691,
+      "step": 6379
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001403194097134005,
+      "loss": 0.7505,
+      "step": 6380
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014030276312588358,
+      "loss": 0.7274,
+      "step": 6381
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014028611520484641,
+      "loss": 0.6194,
+      "step": 6382
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014026946595083995,
+      "loss": 0.7521,
+      "step": 6383
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014025281536441498,
+      "loss": 0.7444,
+      "step": 6384
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014023616344612246,
+      "loss": 0.6965,
+      "step": 6385
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014021951019651334,
+      "loss": 0.6863,
+      "step": 6386
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014020285561613867,
+      "loss": 0.7191,
+      "step": 6387
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001401861997055495,
+      "loss": 0.699,
+      "step": 6388
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014016954246529696,
+      "loss": 0.743,
+      "step": 6389
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014015288389593216,
+      "loss": 0.7211,
+      "step": 6390
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014013622399800627,
+      "loss": 0.748,
+      "step": 6391
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014011956277207058,
+      "loss": 0.7482,
+      "step": 6392
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014010290021867634,
+      "loss": 0.6866,
+      "step": 6393
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001400862363383749,
+      "loss": 0.753,
+      "step": 6394
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001400695711317176,
+      "loss": 0.6556,
+      "step": 6395
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014005290459925585,
+      "loss": 0.6926,
+      "step": 6396
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001400362367415411,
+      "loss": 0.7577,
+      "step": 6397
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00014001956755912488,
+      "loss": 0.6828,
+      "step": 6398
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001400028970525587,
+      "loss": 0.7035,
+      "step": 6399
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013998622522239416,
+      "loss": 0.724,
+      "step": 6400
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001399695520691829,
+      "loss": 0.6696,
+      "step": 6401
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013995287759347659,
+      "loss": 0.6745,
+      "step": 6402
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001399362017958269,
+      "loss": 0.6712,
+      "step": 6403
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013991952467678567,
+      "loss": 0.7772,
+      "step": 6404
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013990284623690465,
+      "loss": 0.7559,
+      "step": 6405
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013988616647673573,
+      "loss": 0.6934,
+      "step": 6406
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0001398694853968308,
+      "loss": 0.7459,
+      "step": 6407
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013985280299774172,
+      "loss": 0.7313,
+      "step": 6408
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013983611928002057,
+      "loss": 0.6485,
+      "step": 6409
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013981943424421932,
+      "loss": 0.7169,
+      "step": 6410
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013980274789089,
+      "loss": 0.746,
+      "step": 6411
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013978606022058482,
+      "loss": 0.6676,
+      "step": 6412
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013976937123385587,
+      "loss": 0.6763,
+      "step": 6413
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00013975268093125538,
+      "loss": 0.6971,
+      "step": 6414
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013973598931333553,
+      "loss": 0.6646,
+      "step": 6415
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013971929638064866,
+      "loss": 0.7225,
+      "step": 6416
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013970260213374712,
+      "loss": 0.6756,
+      "step": 6417
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001396859065731832,
+      "loss": 0.5922,
+      "step": 6418
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013966920969950942,
+      "loss": 0.6852,
+      "step": 6419
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013965251151327812,
+      "loss": 0.7009,
+      "step": 6420
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001396358120150419,
+      "loss": 0.6908,
+      "step": 6421
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013961911120535325,
+      "loss": 0.7019,
+      "step": 6422
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001396024090847648,
+      "loss": 0.7245,
+      "step": 6423
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013958570565382914,
+      "loss": 0.7202,
+      "step": 6424
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013956900091309895,
+      "loss": 0.7273,
+      "step": 6425
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.000139552294863127,
+      "loss": 0.7035,
+      "step": 6426
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013953558750446601,
+      "loss": 0.7346,
+      "step": 6427
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013951887883766878,
+      "loss": 0.6666,
+      "step": 6428
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001395021688632882,
+      "loss": 0.7129,
+      "step": 6429
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001394854575818771,
+      "loss": 0.6569,
+      "step": 6430
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001394687449939885,
+      "loss": 0.636,
+      "step": 6431
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013945203110017526,
+      "loss": 0.7719,
+      "step": 6432
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013943531590099052,
+      "loss": 0.7074,
+      "step": 6433
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001394185993969873,
+      "loss": 0.7271,
+      "step": 6434
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013940188158871868,
+      "loss": 0.7257,
+      "step": 6435
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013938516247673786,
+      "loss": 0.6833,
+      "step": 6436
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013936844206159798,
+      "loss": 0.7109,
+      "step": 6437
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013935172034385234,
+      "loss": 0.6897,
+      "step": 6438
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013933499732405417,
+      "loss": 0.7138,
+      "step": 6439
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013931827300275683,
+      "loss": 0.6897,
+      "step": 6440
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013930154738051365,
+      "loss": 0.749,
+      "step": 6441
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001392848204578781,
+      "loss": 0.711,
+      "step": 6442
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013926809223540357,
+      "loss": 0.6961,
+      "step": 6443
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013925136271364358,
+      "loss": 0.707,
+      "step": 6444
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013923463189315167,
+      "loss": 0.6495,
+      "step": 6445
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013921789977448142,
+      "loss": 0.7342,
+      "step": 6446
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001392011663581865,
+      "loss": 0.7173,
+      "step": 6447
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013918443164482046,
+      "loss": 0.6757,
+      "step": 6448
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001391676956349371,
+      "loss": 0.7175,
+      "step": 6449
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001391509583290902,
+      "loss": 0.7372,
+      "step": 6450
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001391342197278335,
+      "loss": 0.7303,
+      "step": 6451
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013911747983172084,
+      "loss": 0.6366,
+      "step": 6452
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013910073864130613,
+      "loss": 0.6337,
+      "step": 6453
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001390839961571433,
+      "loss": 0.7604,
+      "step": 6454
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013906725237978625,
+      "loss": 0.6695,
+      "step": 6455
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00013905050730978906,
+      "loss": 0.787,
+      "step": 6456
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0001390337609477058,
+      "loss": 0.6449,
+      "step": 6457
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001390170132940905,
+      "loss": 0.6771,
+      "step": 6458
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013900026434949736,
+      "loss": 0.7235,
+      "step": 6459
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001389835141144805,
+      "loss": 0.724,
+      "step": 6460
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013896676258959414,
+      "loss": 0.6564,
+      "step": 6461
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013895000977539267,
+      "loss": 0.6984,
+      "step": 6462
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001389332556724303,
+      "loss": 0.6874,
+      "step": 6463
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013891650028126135,
+      "loss": 0.7356,
+      "step": 6464
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013889974360244028,
+      "loss": 0.683,
+      "step": 6465
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001388829856365215,
+      "loss": 0.7072,
+      "step": 6466
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013886622638405952,
+      "loss": 0.727,
+      "step": 6467
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013884946584560883,
+      "loss": 0.676,
+      "step": 6468
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013883270402172403,
+      "loss": 0.6361,
+      "step": 6469
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001388159409129597,
+      "loss": 0.6659,
+      "step": 6470
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013879917651987047,
+      "loss": 0.7259,
+      "step": 6471
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013878241084301103,
+      "loss": 0.6515,
+      "step": 6472
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001387656438829362,
+      "loss": 0.6687,
+      "step": 6473
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001387488756402007,
+      "loss": 0.6862,
+      "step": 6474
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013873210611535934,
+      "loss": 0.6795,
+      "step": 6475
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013871533530896698,
+      "loss": 0.6883,
+      "step": 6476
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013869856322157854,
+      "loss": 0.7458,
+      "step": 6477
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013868178985374898,
+      "loss": 0.7555,
+      "step": 6478
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001386650152060333,
+      "loss": 0.6893,
+      "step": 6479
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013864823927898647,
+      "loss": 0.7643,
+      "step": 6480
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013863146207316365,
+      "loss": 0.6994,
+      "step": 6481
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013861468358911986,
+      "loss": 0.7519,
+      "step": 6482
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001385979038274103,
+      "loss": 0.722,
+      "step": 6483
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001385811227885902,
+      "loss": 0.6838,
+      "step": 6484
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013856434047321482,
+      "loss": 0.72,
+      "step": 6485
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001385475568818394,
+      "loss": 0.7385,
+      "step": 6486
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013853077201501929,
+      "loss": 0.6936,
+      "step": 6487
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013851398587330982,
+      "loss": 0.6713,
+      "step": 6488
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013849719845726642,
+      "loss": 0.7097,
+      "step": 6489
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013848040976744457,
+      "loss": 0.7076,
+      "step": 6490
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013846361980439977,
+      "loss": 0.7383,
+      "step": 6491
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001384468285686875,
+      "loss": 0.6443,
+      "step": 6492
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013843003606086343,
+      "loss": 0.6773,
+      "step": 6493
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001384132422814831,
+      "loss": 0.703,
+      "step": 6494
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013839644723110224,
+      "loss": 0.7536,
+      "step": 6495
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013837965091027655,
+      "loss": 0.7157,
+      "step": 6496
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0001383628533195617,
+      "loss": 0.7132,
+      "step": 6497
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013834605445951355,
+      "loss": 0.7117,
+      "step": 6498
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013832925433068796,
+      "loss": 0.6955,
+      "step": 6499
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00013831245293364073,
+      "loss": 0.7903,
+      "step": 6500
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013829565026892785,
+      "loss": 0.7171,
+      "step": 6501
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013827884633710525,
+      "loss": 0.674,
+      "step": 6502
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001382620411387289,
+      "loss": 0.7209,
+      "step": 6503
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013824523467435487,
+      "loss": 0.7512,
+      "step": 6504
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013822842694453924,
+      "loss": 0.7433,
+      "step": 6505
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013821161794983815,
+      "loss": 0.668,
+      "step": 6506
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013819480769080778,
+      "loss": 0.6601,
+      "step": 6507
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013817799616800428,
+      "loss": 0.7044,
+      "step": 6508
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013816118338198395,
+      "loss": 0.7104,
+      "step": 6509
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001381443693333031,
+      "loss": 0.7318,
+      "step": 6510
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013812755402251804,
+      "loss": 0.7035,
+      "step": 6511
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013811073745018514,
+      "loss": 0.7458,
+      "step": 6512
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013809391961686082,
+      "loss": 0.6667,
+      "step": 6513
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013807710052310155,
+      "loss": 0.6898,
+      "step": 6514
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013806028016946384,
+      "loss": 0.6943,
+      "step": 6515
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013804345855650423,
+      "loss": 0.7073,
+      "step": 6516
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013802663568477928,
+      "loss": 0.6951,
+      "step": 6517
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001380098115548457,
+      "loss": 0.7086,
+      "step": 6518
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013799298616726005,
+      "loss": 0.6936,
+      "step": 6519
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013797615952257911,
+      "loss": 0.6563,
+      "step": 6520
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001379593316213596,
+      "loss": 0.7246,
+      "step": 6521
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013794250246415836,
+      "loss": 0.6885,
+      "step": 6522
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013792567205153216,
+      "loss": 0.7131,
+      "step": 6523
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013790884038403795,
+      "loss": 0.7243,
+      "step": 6524
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013789200746223257,
+      "loss": 0.7704,
+      "step": 6525
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013787517328667305,
+      "loss": 0.6999,
+      "step": 6526
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013785833785791635,
+      "loss": 0.7372,
+      "step": 6527
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013784150117651954,
+      "loss": 0.6791,
+      "step": 6528
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013782466324303969,
+      "loss": 0.7332,
+      "step": 6529
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013780782405803393,
+      "loss": 0.7584,
+      "step": 6530
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001377909836220594,
+      "loss": 0.6994,
+      "step": 6531
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013777414193567338,
+      "loss": 0.731,
+      "step": 6532
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.000137757298999433,
+      "loss": 0.6975,
+      "step": 6533
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013774045481389568,
+      "loss": 0.7852,
+      "step": 6534
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001377236093796187,
+      "loss": 0.7259,
+      "step": 6535
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013770676269715942,
+      "loss": 0.7221,
+      "step": 6536
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013768991476707527,
+      "loss": 0.7406,
+      "step": 6537
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013767306558992373,
+      "loss": 0.7038,
+      "step": 6538
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013765621516626225,
+      "loss": 0.664,
+      "step": 6539
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013763936349664836,
+      "loss": 0.6767,
+      "step": 6540
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00013762251058163972,
+      "loss": 0.7362,
+      "step": 6541
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001376056564217939,
+      "loss": 0.6862,
+      "step": 6542
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001375888010176686,
+      "loss": 0.6897,
+      "step": 6543
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013757194436982143,
+      "loss": 0.73,
+      "step": 6544
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013755508647881023,
+      "loss": 0.6905,
+      "step": 6545
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013753822734519276,
+      "loss": 0.7361,
+      "step": 6546
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013752136696952683,
+      "loss": 0.7105,
+      "step": 6547
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013750450535237037,
+      "loss": 0.7245,
+      "step": 6548
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0001374876424942812,
+      "loss": 0.6723,
+      "step": 6549
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013747077839581728,
+      "loss": 0.6615,
+      "step": 6550
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013745391305753667,
+      "loss": 0.7073,
+      "step": 6551
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013743704647999736,
+      "loss": 0.6378,
+      "step": 6552
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013742017866375745,
+      "loss": 0.715,
+      "step": 6553
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013740330960937503,
+      "loss": 0.6374,
+      "step": 6554
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013738643931740827,
+      "loss": 0.6596,
+      "step": 6555
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013736956778841533,
+      "loss": 0.6785,
+      "step": 6556
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013735269502295448,
+      "loss": 0.6934,
+      "step": 6557
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.000137335821021584,
+      "loss": 0.7343,
+      "step": 6558
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013731894578486225,
+      "loss": 0.6768,
+      "step": 6559
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0001373020693133475,
+      "loss": 0.7113,
+      "step": 6560
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013728519160759818,
+      "loss": 0.6876,
+      "step": 6561
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013726831266817278,
+      "loss": 0.7046,
+      "step": 6562
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013725143249562974,
+      "loss": 0.6961,
+      "step": 6563
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013723455109052762,
+      "loss": 0.7006,
+      "step": 6564
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013721766845342493,
+      "loss": 0.66,
+      "step": 6565
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013720078458488036,
+      "loss": 0.6749,
+      "step": 6566
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0001371838994854524,
+      "loss": 0.6167,
+      "step": 6567
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013716701315569991,
+      "loss": 0.6532,
+      "step": 6568
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013715012559618154,
+      "loss": 0.7328,
+      "step": 6569
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013713323680745608,
+      "loss": 0.7619,
+      "step": 6570
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0001371163467900823,
+      "loss": 0.7625,
+      "step": 6571
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013709945554461907,
+      "loss": 0.7354,
+      "step": 6572
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013708256307162525,
+      "loss": 0.7318,
+      "step": 6573
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013706566937165984,
+      "loss": 0.7909,
+      "step": 6574
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013704877444528182,
+      "loss": 0.7229,
+      "step": 6575
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0001370318782930501,
+      "loss": 0.7366,
+      "step": 6576
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0001370149809155238,
+      "loss": 0.7357,
+      "step": 6577
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.000136998082313262,
+      "loss": 0.6983,
+      "step": 6578
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013698118248682382,
+      "loss": 0.6655,
+      "step": 6579
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0001369642814367685,
+      "loss": 0.7358,
+      "step": 6580
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013694737916365517,
+      "loss": 0.6901,
+      "step": 6581
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0001369304756680431,
+      "loss": 0.6684,
+      "step": 6582
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013691357095049165,
+      "loss": 0.6977,
+      "step": 6583
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013689666501156005,
+      "loss": 0.675,
+      "step": 6584
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013687975785180777,
+      "loss": 0.8054,
+      "step": 6585
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00013686284947179425,
+      "loss": 0.6745,
+      "step": 6586
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013684593987207881,
+      "loss": 0.7315,
+      "step": 6587
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013682902905322105,
+      "loss": 0.7315,
+      "step": 6588
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013681211701578052,
+      "loss": 0.7207,
+      "step": 6589
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013679520376031672,
+      "loss": 0.6735,
+      "step": 6590
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013677828928738934,
+      "loss": 0.6796,
+      "step": 6591
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.000136761373597558,
+      "loss": 0.7345,
+      "step": 6592
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013674445669138241,
+      "loss": 0.7423,
+      "step": 6593
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001367275385694223,
+      "loss": 0.7169,
+      "step": 6594
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013671061923223746,
+      "loss": 0.6732,
+      "step": 6595
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013669369868038773,
+      "loss": 0.8212,
+      "step": 6596
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013667677691443295,
+      "loss": 0.6889,
+      "step": 6597
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013665985393493299,
+      "loss": 0.6718,
+      "step": 6598
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013664292974244782,
+      "loss": 0.6635,
+      "step": 6599
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013662600433753745,
+      "loss": 0.691,
+      "step": 6600
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013660907772076187,
+      "loss": 0.7796,
+      "step": 6601
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001365921498926811,
+      "loss": 0.688,
+      "step": 6602
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001365752208538553,
+      "loss": 0.6879,
+      "step": 6603
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001365582906048446,
+      "loss": 0.7587,
+      "step": 6604
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013654135914620917,
+      "loss": 0.7267,
+      "step": 6605
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001365244264785092,
+      "loss": 0.6962,
+      "step": 6606
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013650749260230502,
+      "loss": 0.6695,
+      "step": 6607
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001364905575181569,
+      "loss": 0.769,
+      "step": 6608
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013647362122662515,
+      "loss": 0.6652,
+      "step": 6609
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013645668372827015,
+      "loss": 0.6888,
+      "step": 6610
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001364397450236524,
+      "loss": 0.6454,
+      "step": 6611
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001364228051133323,
+      "loss": 0.6537,
+      "step": 6612
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001364058639978703,
+      "loss": 0.6512,
+      "step": 6613
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013638892167782703,
+      "loss": 0.7281,
+      "step": 6614
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013637197815376305,
+      "loss": 0.761,
+      "step": 6615
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013635503342623897,
+      "loss": 0.704,
+      "step": 6616
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013633808749581542,
+      "loss": 0.74,
+      "step": 6617
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013632114036305314,
+      "loss": 0.6907,
+      "step": 6618
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013630419202851284,
+      "loss": 0.6816,
+      "step": 6619
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013628724249275536,
+      "loss": 0.6542,
+      "step": 6620
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013627029175634144,
+      "loss": 0.695,
+      "step": 6621
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013625333981983194,
+      "loss": 0.7283,
+      "step": 6622
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013623638668378783,
+      "loss": 0.7441,
+      "step": 6623
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013621943234877,
+      "loss": 0.7719,
+      "step": 6624
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013620247681533942,
+      "loss": 0.7227,
+      "step": 6625
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013618552008405713,
+      "loss": 0.6764,
+      "step": 6626
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013616856215548418,
+      "loss": 0.6717,
+      "step": 6627
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001361516030301817,
+      "loss": 0.6907,
+      "step": 6628
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013613464270871076,
+      "loss": 0.7002,
+      "step": 6629
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00013611768119163259,
+      "loss": 0.7168,
+      "step": 6630
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013610071847950834,
+      "loss": 0.6083,
+      "step": 6631
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001360837545728993,
+      "loss": 0.7579,
+      "step": 6632
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013606678947236686,
+      "loss": 0.7154,
+      "step": 6633
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001360498231784722,
+      "loss": 0.7349,
+      "step": 6634
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013603285569177675,
+      "loss": 0.6976,
+      "step": 6635
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013601588701284197,
+      "loss": 0.6974,
+      "step": 6636
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013599891714222924,
+      "loss": 0.6305,
+      "step": 6637
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001359819460805001,
+      "loss": 0.7136,
+      "step": 6638
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001359649738282161,
+      "loss": 0.7169,
+      "step": 6639
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013594800038593872,
+      "loss": 0.6983,
+      "step": 6640
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013593102575422964,
+      "loss": 0.6931,
+      "step": 6641
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001359140499336505,
+      "loss": 0.7623,
+      "step": 6642
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013589707292476302,
+      "loss": 0.7429,
+      "step": 6643
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013588009472812889,
+      "loss": 0.7415,
+      "step": 6644
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013586311534430982,
+      "loss": 0.7145,
+      "step": 6645
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013584613477386773,
+      "loss": 0.6447,
+      "step": 6646
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001358291530173644,
+      "loss": 0.7238,
+      "step": 6647
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013581217007536172,
+      "loss": 0.7002,
+      "step": 6648
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013579518594842161,
+      "loss": 0.6764,
+      "step": 6649
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013577820063710608,
+      "loss": 0.6509,
+      "step": 6650
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013576121414197705,
+      "loss": 0.6504,
+      "step": 6651
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013574422646359663,
+      "loss": 0.7739,
+      "step": 6652
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013572723760252688,
+      "loss": 0.7503,
+      "step": 6653
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013571024755932993,
+      "loss": 0.6799,
+      "step": 6654
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013569325633456793,
+      "loss": 0.6731,
+      "step": 6655
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001356762639288031,
+      "loss": 0.7218,
+      "step": 6656
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001356592703425976,
+      "loss": 0.7584,
+      "step": 6657
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013564227557651376,
+      "loss": 0.6731,
+      "step": 6658
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013562527963111392,
+      "loss": 0.6495,
+      "step": 6659
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013560828250696043,
+      "loss": 0.7048,
+      "step": 6660
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013559128420461564,
+      "loss": 0.7315,
+      "step": 6661
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.000135574284724642,
+      "loss": 0.7057,
+      "step": 6662
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.000135557284067602,
+      "loss": 0.6461,
+      "step": 6663
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001355402822340581,
+      "loss": 0.7147,
+      "step": 6664
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013552327922457298,
+      "loss": 0.6975,
+      "step": 6665
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013550627503970908,
+      "loss": 0.7346,
+      "step": 6666
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013548926968002907,
+      "loss": 0.7197,
+      "step": 6667
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013547226314609563,
+      "loss": 0.761,
+      "step": 6668
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013545525543847146,
+      "loss": 0.709,
+      "step": 6669
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0001354382465577193,
+      "loss": 0.7637,
+      "step": 6670
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.000135421236504402,
+      "loss": 0.7541,
+      "step": 6671
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013540422527908227,
+      "loss": 0.7743,
+      "step": 6672
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00013538721288232304,
+      "loss": 0.6866,
+      "step": 6673
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013537019931468715,
+      "loss": 0.6862,
+      "step": 6674
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001353531845767376,
+      "loss": 0.6702,
+      "step": 6675
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013533616866903735,
+      "loss": 0.7074,
+      "step": 6676
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001353191515921494,
+      "loss": 0.6624,
+      "step": 6677
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013530213334663683,
+      "loss": 0.7291,
+      "step": 6678
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013528511393306268,
+      "loss": 0.799,
+      "step": 6679
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013526809335199014,
+      "loss": 0.7151,
+      "step": 6680
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013525107160398237,
+      "loss": 0.6973,
+      "step": 6681
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013523404868960253,
+      "loss": 0.6784,
+      "step": 6682
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001352170246094139,
+      "loss": 0.7246,
+      "step": 6683
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013519999936397974,
+      "loss": 0.6847,
+      "step": 6684
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013518297295386344,
+      "loss": 0.7443,
+      "step": 6685
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001351659453796283,
+      "loss": 0.7006,
+      "step": 6686
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013514891664183777,
+      "loss": 0.6812,
+      "step": 6687
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013513188674105524,
+      "loss": 0.7732,
+      "step": 6688
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001351148556778442,
+      "loss": 0.6404,
+      "step": 6689
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013509782345276817,
+      "loss": 0.6945,
+      "step": 6690
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013508079006639072,
+      "loss": 0.7154,
+      "step": 6691
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013506375551927547,
+      "loss": 0.7429,
+      "step": 6692
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013504671981198594,
+      "loss": 0.6607,
+      "step": 6693
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013502968294508592,
+      "loss": 0.6583,
+      "step": 6694
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013501264491913906,
+      "loss": 0.7211,
+      "step": 6695
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013499560573470914,
+      "loss": 0.8086,
+      "step": 6696
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013497856539235995,
+      "loss": 0.7364,
+      "step": 6697
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013496152389265523,
+      "loss": 0.6505,
+      "step": 6698
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013494448123615893,
+      "loss": 0.6882,
+      "step": 6699
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001349274374234349,
+      "loss": 0.6356,
+      "step": 6700
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001349103924550471,
+      "loss": 0.6603,
+      "step": 6701
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001348933463315595,
+      "loss": 0.7258,
+      "step": 6702
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013487629905353617,
+      "loss": 0.7595,
+      "step": 6703
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013485925062154105,
+      "loss": 0.6694,
+      "step": 6704
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013484220103613832,
+      "loss": 0.721,
+      "step": 6705
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001348251502978921,
+      "loss": 0.7469,
+      "step": 6706
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001348080984073665,
+      "loss": 0.683,
+      "step": 6707
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001347910453651258,
+      "loss": 0.6913,
+      "step": 6708
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013477399117173418,
+      "loss": 0.6743,
+      "step": 6709
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013475693582775598,
+      "loss": 0.7008,
+      "step": 6710
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013473987933375548,
+      "loss": 0.751,
+      "step": 6711
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013472282169029704,
+      "loss": 0.693,
+      "step": 6712
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013470576289794506,
+      "loss": 0.6732,
+      "step": 6713
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013468870295726398,
+      "loss": 0.738,
+      "step": 6714
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0001346716418688183,
+      "loss": 0.6784,
+      "step": 6715
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00013465457963317246,
+      "loss": 0.6833,
+      "step": 6716
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013463751625089105,
+      "loss": 0.6664,
+      "step": 6717
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013462045172253867,
+      "loss": 0.7264,
+      "step": 6718
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013460338604867993,
+      "loss": 0.6449,
+      "step": 6719
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013458631922987947,
+      "loss": 0.6842,
+      "step": 6720
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013456925126670204,
+      "loss": 0.7134,
+      "step": 6721
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001345521821597123,
+      "loss": 0.7096,
+      "step": 6722
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001345351119094751,
+      "loss": 0.7149,
+      "step": 6723
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013451804051655523,
+      "loss": 0.7178,
+      "step": 6724
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001345009679815175,
+      "loss": 0.8252,
+      "step": 6725
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013448389430492687,
+      "loss": 0.7466,
+      "step": 6726
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001344668194873482,
+      "loss": 0.7954,
+      "step": 6727
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001344497435293465,
+      "loss": 0.6349,
+      "step": 6728
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013443266643148674,
+      "loss": 0.7394,
+      "step": 6729
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013441558819433398,
+      "loss": 0.6999,
+      "step": 6730
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001343985088184533,
+      "loss": 0.6535,
+      "step": 6731
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001343814283044098,
+      "loss": 0.652,
+      "step": 6732
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013436434665276865,
+      "loss": 0.6617,
+      "step": 6733
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013434726386409504,
+      "loss": 0.6516,
+      "step": 6734
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013433017993895417,
+      "loss": 0.6914,
+      "step": 6735
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001343130948779113,
+      "loss": 0.7524,
+      "step": 6736
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001342960086815318,
+      "loss": 0.7061,
+      "step": 6737
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001342789213503809,
+      "loss": 0.6328,
+      "step": 6738
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001342618328850241,
+      "loss": 0.6948,
+      "step": 6739
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013424474328602677,
+      "loss": 0.6882,
+      "step": 6740
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013422765255395431,
+      "loss": 0.6476,
+      "step": 6741
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013421056068937227,
+      "loss": 0.6883,
+      "step": 6742
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013419346769284616,
+      "loss": 0.6902,
+      "step": 6743
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013417637356494155,
+      "loss": 0.6982,
+      "step": 6744
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013415927830622408,
+      "loss": 0.6813,
+      "step": 6745
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001341421819172593,
+      "loss": 0.6758,
+      "step": 6746
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013412508439861293,
+      "loss": 0.6375,
+      "step": 6747
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013410798575085072,
+      "loss": 0.7294,
+      "step": 6748
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001340908859745384,
+      "loss": 0.7719,
+      "step": 6749
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013407378507024177,
+      "loss": 0.6911,
+      "step": 6750
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013405668303852664,
+      "loss": 0.7515,
+      "step": 6751
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013403957987995882,
+      "loss": 0.6601,
+      "step": 6752
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001340224755951043,
+      "loss": 0.7163,
+      "step": 6753
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.000134005370184529,
+      "loss": 0.7132,
+      "step": 6754
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001339882636487989,
+      "loss": 0.6744,
+      "step": 6755
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013397115598848004,
+      "loss": 0.6848,
+      "step": 6756
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013395404720413834,
+      "loss": 0.7531,
+      "step": 6757
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013393693729634002,
+      "loss": 0.7381,
+      "step": 6758
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00013391982626565114,
+      "loss": 0.7334,
+      "step": 6759
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001339027141126379,
+      "loss": 0.661,
+      "step": 6760
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013388560083786653,
+      "loss": 0.7552,
+      "step": 6761
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013386848644190318,
+      "loss": 0.6651,
+      "step": 6762
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013385137092531413,
+      "loss": 0.7383,
+      "step": 6763
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013383425428866577,
+      "loss": 0.606,
+      "step": 6764
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013381713653252438,
+      "loss": 0.6578,
+      "step": 6765
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001338000176574564,
+      "loss": 0.6638,
+      "step": 6766
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001337828976640282,
+      "loss": 0.7462,
+      "step": 6767
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013376577655280626,
+      "loss": 0.6608,
+      "step": 6768
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013374865432435706,
+      "loss": 0.6931,
+      "step": 6769
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013373153097924717,
+      "loss": 0.713,
+      "step": 6770
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013371440651804313,
+      "loss": 0.6551,
+      "step": 6771
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001336972809413116,
+      "loss": 0.65,
+      "step": 6772
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001336801542496191,
+      "loss": 0.6926,
+      "step": 6773
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013366302644353243,
+      "loss": 0.7106,
+      "step": 6774
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013364589752361822,
+      "loss": 0.7121,
+      "step": 6775
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013362876749044333,
+      "loss": 0.709,
+      "step": 6776
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013361163634457448,
+      "loss": 0.7622,
+      "step": 6777
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001335945040865785,
+      "loss": 0.7049,
+      "step": 6778
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013357737071702225,
+      "loss": 0.7634,
+      "step": 6779
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013356023623647265,
+      "loss": 0.6703,
+      "step": 6780
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013354310064549662,
+      "loss": 0.6673,
+      "step": 6781
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013352596394466119,
+      "loss": 0.7178,
+      "step": 6782
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001335088261345333,
+      "loss": 0.7506,
+      "step": 6783
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013349168721568005,
+      "loss": 0.7099,
+      "step": 6784
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013347454718866847,
+      "loss": 0.7226,
+      "step": 6785
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013345740605406572,
+      "loss": 0.6705,
+      "step": 6786
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013344026381243896,
+      "loss": 0.6955,
+      "step": 6787
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013342312046435542,
+      "loss": 0.7779,
+      "step": 6788
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013340597601038223,
+      "loss": 0.7194,
+      "step": 6789
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013338883045108674,
+      "loss": 0.6829,
+      "step": 6790
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013337168378703624,
+      "loss": 0.6837,
+      "step": 6791
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013335453601879803,
+      "loss": 0.6805,
+      "step": 6792
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013333738714693956,
+      "loss": 0.7422,
+      "step": 6793
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001333202371720282,
+      "loss": 0.6753,
+      "step": 6794
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0001333030860946314,
+      "loss": 0.6927,
+      "step": 6795
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013328593391531666,
+      "loss": 0.7417,
+      "step": 6796
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013326878063465148,
+      "loss": 0.7522,
+      "step": 6797
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013325162625320346,
+      "loss": 0.7098,
+      "step": 6798
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013323447077154016,
+      "loss": 0.7056,
+      "step": 6799
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013321731419022924,
+      "loss": 0.7238,
+      "step": 6800
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013320015650983836,
+      "loss": 0.7119,
+      "step": 6801
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00013318299773093522,
+      "loss": 0.7194,
+      "step": 6802
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013316583785408755,
+      "loss": 0.7378,
+      "step": 6803
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013314867687986317,
+      "loss": 0.7521,
+      "step": 6804
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013313151480882985,
+      "loss": 0.6933,
+      "step": 6805
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013311435164155546,
+      "loss": 0.6917,
+      "step": 6806
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001330971873786079,
+      "loss": 0.6962,
+      "step": 6807
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013308002202055508,
+      "loss": 0.6831,
+      "step": 6808
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013306285556796495,
+      "loss": 0.7213,
+      "step": 6809
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001330456880214055,
+      "loss": 0.7221,
+      "step": 6810
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001330285193814448,
+      "loss": 0.6712,
+      "step": 6811
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013301134964865092,
+      "loss": 0.6188,
+      "step": 6812
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013299417882359191,
+      "loss": 0.7011,
+      "step": 6813
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013297700690683597,
+      "loss": 0.6598,
+      "step": 6814
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013295983389895123,
+      "loss": 0.6346,
+      "step": 6815
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013294265980050594,
+      "loss": 0.7412,
+      "step": 6816
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001329254846120683,
+      "loss": 0.7053,
+      "step": 6817
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013290830833420663,
+      "loss": 0.6958,
+      "step": 6818
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001328911309674893,
+      "loss": 0.7649,
+      "step": 6819
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013287395251248456,
+      "loss": 0.6567,
+      "step": 6820
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013285677296976086,
+      "loss": 0.7409,
+      "step": 6821
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001328395923398867,
+      "loss": 0.698,
+      "step": 6822
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013282241062343038,
+      "loss": 0.7526,
+      "step": 6823
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013280522782096056,
+      "loss": 0.712,
+      "step": 6824
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013278804393304568,
+      "loss": 0.7222,
+      "step": 6825
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013277085896025434,
+      "loss": 0.6763,
+      "step": 6826
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013275367290315515,
+      "loss": 0.7,
+      "step": 6827
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001327364857623168,
+      "loss": 0.6663,
+      "step": 6828
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013271929753830788,
+      "loss": 0.7383,
+      "step": 6829
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013270210823169718,
+      "loss": 0.7344,
+      "step": 6830
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013268491784305342,
+      "loss": 0.7569,
+      "step": 6831
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013266772637294538,
+      "loss": 0.6886,
+      "step": 6832
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013265053382194192,
+      "loss": 0.6704,
+      "step": 6833
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013263334019061188,
+      "loss": 0.7012,
+      "step": 6834
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013261614547952416,
+      "loss": 0.7037,
+      "step": 6835
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013259894968924765,
+      "loss": 0.741,
+      "step": 6836
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013258175282035136,
+      "loss": 0.7313,
+      "step": 6837
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001325645548734043,
+      "loss": 0.6861,
+      "step": 6838
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001325473558489755,
+      "loss": 0.7494,
+      "step": 6839
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.000132530155747634,
+      "loss": 0.6867,
+      "step": 6840
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013251295456994899,
+      "loss": 0.6676,
+      "step": 6841
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013249575231648948,
+      "loss": 0.7415,
+      "step": 6842
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001324785489878248,
+      "loss": 0.6957,
+      "step": 6843
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013246134458452406,
+      "loss": 0.7124,
+      "step": 6844
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00013244413910715657,
+      "loss": 0.7,
+      "step": 6845
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0001324269325562916,
+      "loss": 0.6121,
+      "step": 6846
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013240972493249847,
+      "loss": 0.7372,
+      "step": 6847
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001323925162363465,
+      "loss": 0.6477,
+      "step": 6848
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013237530646840516,
+      "loss": 0.7254,
+      "step": 6849
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001323580956292438,
+      "loss": 0.6822,
+      "step": 6850
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.000132340883719432,
+      "loss": 0.7156,
+      "step": 6851
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013232367073953913,
+      "loss": 0.7198,
+      "step": 6852
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013230645669013474,
+      "loss": 0.6114,
+      "step": 6853
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001322892415717885,
+      "loss": 0.7346,
+      "step": 6854
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013227202538506995,
+      "loss": 0.7426,
+      "step": 6855
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013225480813054875,
+      "loss": 0.7751,
+      "step": 6856
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013223758980879458,
+      "loss": 0.6869,
+      "step": 6857
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001322203704203771,
+      "loss": 0.6742,
+      "step": 6858
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013220314996586612,
+      "loss": 0.7986,
+      "step": 6859
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001321859284458314,
+      "loss": 0.7208,
+      "step": 6860
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013216870586084273,
+      "loss": 0.6832,
+      "step": 6861
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013215148221147006,
+      "loss": 0.6368,
+      "step": 6862
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013213425749828315,
+      "loss": 0.6553,
+      "step": 6863
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.000132117031721852,
+      "loss": 0.71,
+      "step": 6864
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013209980488274652,
+      "loss": 0.6994,
+      "step": 6865
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013208257698153677,
+      "loss": 0.6191,
+      "step": 6866
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013206534801879274,
+      "loss": 0.696,
+      "step": 6867
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013204811799508452,
+      "loss": 0.738,
+      "step": 6868
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013203088691098218,
+      "loss": 0.6921,
+      "step": 6869
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013201365476705584,
+      "loss": 0.6871,
+      "step": 6870
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013199642156387571,
+      "loss": 0.6286,
+      "step": 6871
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013197918730201199,
+      "loss": 0.7449,
+      "step": 6872
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001319619519820349,
+      "loss": 0.7434,
+      "step": 6873
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013194471560451474,
+      "loss": 0.694,
+      "step": 6874
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013192747817002178,
+      "loss": 0.6858,
+      "step": 6875
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013191023967912637,
+      "loss": 0.6548,
+      "step": 6876
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013189300013239893,
+      "loss": 0.6949,
+      "step": 6877
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013187575953040986,
+      "loss": 0.72,
+      "step": 6878
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001318585178737296,
+      "loss": 0.7472,
+      "step": 6879
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013184127516292863,
+      "loss": 0.7234,
+      "step": 6880
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013182403139857748,
+      "loss": 0.7491,
+      "step": 6881
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001318067865812467,
+      "loss": 0.745,
+      "step": 6882
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013178954071150687,
+      "loss": 0.7318,
+      "step": 6883
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001317722937899286,
+      "loss": 0.7131,
+      "step": 6884
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001317550458170826,
+      "loss": 0.676,
+      "step": 6885
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013173779679353956,
+      "loss": 0.6657,
+      "step": 6886
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00013172054671987013,
+      "loss": 0.7332,
+      "step": 6887
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001317032955966451,
+      "loss": 0.7166,
+      "step": 6888
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001316860434244354,
+      "loss": 0.6219,
+      "step": 6889
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001316687902038117,
+      "loss": 0.742,
+      "step": 6890
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001316515359353449,
+      "loss": 0.7268,
+      "step": 6891
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013163428061960594,
+      "loss": 0.6693,
+      "step": 6892
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013161702425716572,
+      "loss": 0.7151,
+      "step": 6893
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013159976684859527,
+      "loss": 0.6886,
+      "step": 6894
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013158250839446554,
+      "loss": 0.6987,
+      "step": 6895
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013156524889534757,
+      "loss": 0.6954,
+      "step": 6896
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013154798835181248,
+      "loss": 0.7579,
+      "step": 6897
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013153072676443133,
+      "loss": 0.697,
+      "step": 6898
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001315134641337753,
+      "loss": 0.7314,
+      "step": 6899
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013149620046041554,
+      "loss": 0.6458,
+      "step": 6900
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013147893574492328,
+      "loss": 0.7058,
+      "step": 6901
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013146166998786974,
+      "loss": 0.6569,
+      "step": 6902
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013144440318982626,
+      "loss": 0.6695,
+      "step": 6903
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013142713535136414,
+      "loss": 0.6704,
+      "step": 6904
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013140986647305466,
+      "loss": 0.6844,
+      "step": 6905
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013139259655546928,
+      "loss": 0.7003,
+      "step": 6906
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013137532559917938,
+      "loss": 0.7574,
+      "step": 6907
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013135805360475647,
+      "loss": 0.6788,
+      "step": 6908
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013134078057277195,
+      "loss": 0.7497,
+      "step": 6909
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013132350650379743,
+      "loss": 0.6636,
+      "step": 6910
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001313062313984044,
+      "loss": 0.7488,
+      "step": 6911
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013128895525716452,
+      "loss": 0.7265,
+      "step": 6912
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013127167808064933,
+      "loss": 0.6438,
+      "step": 6913
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013125439986943056,
+      "loss": 0.6928,
+      "step": 6914
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013123712062407985,
+      "loss": 0.7322,
+      "step": 6915
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013121984034516899,
+      "loss": 0.6584,
+      "step": 6916
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013120255903326968,
+      "loss": 0.6874,
+      "step": 6917
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013118527668895378,
+      "loss": 0.6519,
+      "step": 6918
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013116799331279306,
+      "loss": 0.7051,
+      "step": 6919
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001311507089053594,
+      "loss": 0.7692,
+      "step": 6920
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013113342346722474,
+      "loss": 0.6582,
+      "step": 6921
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013111613699896095,
+      "loss": 0.7115,
+      "step": 6922
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013109884950114007,
+      "loss": 0.668,
+      "step": 6923
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013108156097433403,
+      "loss": 0.6955,
+      "step": 6924
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0001310642714191149,
+      "loss": 0.6811,
+      "step": 6925
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013104698083605474,
+      "loss": 0.6248,
+      "step": 6926
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013102968922572568,
+      "loss": 0.7407,
+      "step": 6927
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013101239658869983,
+      "loss": 0.6874,
+      "step": 6928
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013099510292554934,
+      "loss": 0.7235,
+      "step": 6929
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013097780823684646,
+      "loss": 0.6811,
+      "step": 6930
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013096051252316343,
+      "loss": 0.6956,
+      "step": 6931
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00013094321578507246,
+      "loss": 0.7342,
+      "step": 6932
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013092591802314594,
+      "loss": 0.705,
+      "step": 6933
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013090861923795616,
+      "loss": 0.7044,
+      "step": 6934
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013089131943007552,
+      "loss": 0.644,
+      "step": 6935
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001308740186000764,
+      "loss": 0.7341,
+      "step": 6936
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013085671674853124,
+      "loss": 0.7489,
+      "step": 6937
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013083941387601256,
+      "loss": 0.715,
+      "step": 6938
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013082210998309284,
+      "loss": 0.6571,
+      "step": 6939
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001308048050703446,
+      "loss": 0.7081,
+      "step": 6940
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013078749913834046,
+      "loss": 0.6846,
+      "step": 6941
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013077019218765305,
+      "loss": 0.7169,
+      "step": 6942
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001307528842188549,
+      "loss": 0.7419,
+      "step": 6943
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013073557523251886,
+      "loss": 0.6728,
+      "step": 6944
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001307182652292175,
+      "loss": 0.7046,
+      "step": 6945
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013070095420952363,
+      "loss": 0.7036,
+      "step": 6946
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013068364217401002,
+      "loss": 0.6638,
+      "step": 6947
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013066632912324947,
+      "loss": 0.6644,
+      "step": 6948
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013064901505781485,
+      "loss": 0.6731,
+      "step": 6949
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013063169997827898,
+      "loss": 0.718,
+      "step": 6950
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013061438388521488,
+      "loss": 0.7298,
+      "step": 6951
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001305970667791954,
+      "loss": 0.6753,
+      "step": 6952
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001305797486607936,
+      "loss": 0.7005,
+      "step": 6953
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013056242953058238,
+      "loss": 0.7015,
+      "step": 6954
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001305451093891349,
+      "loss": 0.7011,
+      "step": 6955
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001305277882370242,
+      "loss": 0.7171,
+      "step": 6956
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001305104660748234,
+      "loss": 0.7262,
+      "step": 6957
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013049314290310566,
+      "loss": 0.6371,
+      "step": 6958
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013047581872244413,
+      "loss": 0.6248,
+      "step": 6959
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013045849353341202,
+      "loss": 0.6869,
+      "step": 6960
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001304411673365826,
+      "loss": 0.6605,
+      "step": 6961
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013042384013252917,
+      "loss": 0.6977,
+      "step": 6962
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013040651192182504,
+      "loss": 0.7004,
+      "step": 6963
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013038918270504355,
+      "loss": 0.7677,
+      "step": 6964
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013037185248275803,
+      "loss": 0.7064,
+      "step": 6965
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013035452125554196,
+      "loss": 0.7243,
+      "step": 6966
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013033718902396877,
+      "loss": 0.6944,
+      "step": 6967
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013031985578861196,
+      "loss": 0.7261,
+      "step": 6968
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.000130302521550045,
+      "loss": 0.6677,
+      "step": 6969
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013028518630884146,
+      "loss": 0.667,
+      "step": 6970
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013026785006557492,
+      "loss": 0.6694,
+      "step": 6971
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013025051282081898,
+      "loss": 0.6638,
+      "step": 6972
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013023317457514733,
+      "loss": 0.6582,
+      "step": 6973
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013021583532913362,
+      "loss": 0.7341,
+      "step": 6974
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00013019849508335158,
+      "loss": 0.7203,
+      "step": 6975
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00013018115383837492,
+      "loss": 0.7192,
+      "step": 6976
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00013016381159477745,
+      "loss": 0.7109,
+      "step": 6977
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00013014646835313298,
+      "loss": 0.7422,
+      "step": 6978
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00013012912411401536,
+      "loss": 0.7383,
+      "step": 6979
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00013011177887799845,
+      "loss": 0.7474,
+      "step": 6980
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00013009443264565618,
+      "loss": 0.6814,
+      "step": 6981
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001300770854175625,
+      "loss": 0.7059,
+      "step": 6982
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001300597371942913,
+      "loss": 0.7183,
+      "step": 6983
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00013004238797641675,
+      "loss": 0.719,
+      "step": 6984
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001300250377645128,
+      "loss": 0.712,
+      "step": 6985
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00013000768655915349,
+      "loss": 0.6781,
+      "step": 6986
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.000129990334360913,
+      "loss": 0.6788,
+      "step": 6987
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012997298117036542,
+      "loss": 0.6217,
+      "step": 6988
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012995562698808494,
+      "loss": 0.7409,
+      "step": 6989
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012993827181464584,
+      "loss": 0.678,
+      "step": 6990
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012992091565062225,
+      "loss": 0.7306,
+      "step": 6991
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001299035584965885,
+      "loss": 0.685,
+      "step": 6992
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001298862003531189,
+      "loss": 0.7053,
+      "step": 6993
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012986884122078774,
+      "loss": 0.716,
+      "step": 6994
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012985148110016947,
+      "loss": 0.7042,
+      "step": 6995
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012983411999183844,
+      "loss": 0.7455,
+      "step": 6996
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012981675789636905,
+      "loss": 0.7093,
+      "step": 6997
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001297993948143359,
+      "loss": 0.7263,
+      "step": 6998
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012978203074631334,
+      "loss": 0.6977,
+      "step": 6999
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.000129764665692876,
+      "loss": 0.706,
+      "step": 7000
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012974729965459843,
+      "loss": 0.6838,
+      "step": 7001
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001297299326320552,
+      "loss": 0.6906,
+      "step": 7002
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012971256462582099,
+      "loss": 0.7652,
+      "step": 7003
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012969519563647041,
+      "loss": 0.7305,
+      "step": 7004
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012967782566457818,
+      "loss": 0.6605,
+      "step": 7005
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012966045471071906,
+      "loss": 0.687,
+      "step": 7006
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012964308277546777,
+      "loss": 0.7317,
+      "step": 7007
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012962570985939915,
+      "loss": 0.7773,
+      "step": 7008
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.000129608335963088,
+      "loss": 0.6971,
+      "step": 7009
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012959096108710916,
+      "loss": 0.6669,
+      "step": 7010
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012957358523203754,
+      "loss": 0.6921,
+      "step": 7011
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012955620839844807,
+      "loss": 0.7246,
+      "step": 7012
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001295388305869157,
+      "loss": 0.697,
+      "step": 7013
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012952145179801546,
+      "loss": 0.7711,
+      "step": 7014
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001295040720323223,
+      "loss": 0.7396,
+      "step": 7015
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0001294866912904113,
+      "loss": 0.7198,
+      "step": 7016
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012946930957285759,
+      "loss": 0.729,
+      "step": 7017
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00012945192688023624,
+      "loss": 0.7018,
+      "step": 7018
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012943454321312243,
+      "loss": 0.7546,
+      "step": 7019
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001294171585720913,
+      "loss": 0.7151,
+      "step": 7020
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012939977295771813,
+      "loss": 0.6884,
+      "step": 7021
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012938238637057813,
+      "loss": 0.7435,
+      "step": 7022
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012936499881124655,
+      "loss": 0.6544,
+      "step": 7023
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012934761028029874,
+      "loss": 0.6803,
+      "step": 7024
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012933022077831006,
+      "loss": 0.7155,
+      "step": 7025
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012931283030585583,
+      "loss": 0.6685,
+      "step": 7026
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012929543886351153,
+      "loss": 0.7097,
+      "step": 7027
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012927804645185254,
+      "loss": 0.6971,
+      "step": 7028
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012926065307145436,
+      "loss": 0.7284,
+      "step": 7029
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001292432587228925,
+      "loss": 0.6131,
+      "step": 7030
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001292258634067425,
+      "loss": 0.6172,
+      "step": 7031
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001292084671235799,
+      "loss": 0.6766,
+      "step": 7032
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001291910698739803,
+      "loss": 0.7629,
+      "step": 7033
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012917367165851933,
+      "loss": 0.671,
+      "step": 7034
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012915627247777272,
+      "loss": 0.8079,
+      "step": 7035
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001291388723323161,
+      "loss": 0.7142,
+      "step": 7036
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012912147122272523,
+      "loss": 0.6929,
+      "step": 7037
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012910406914957582,
+      "loss": 0.7206,
+      "step": 7038
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012908666611344374,
+      "loss": 0.6797,
+      "step": 7039
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012906926211490475,
+      "loss": 0.6807,
+      "step": 7040
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012905185715453475,
+      "loss": 0.6661,
+      "step": 7041
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001290344512329096,
+      "loss": 0.7213,
+      "step": 7042
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012901704435060526,
+      "loss": 0.7149,
+      "step": 7043
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001289996365081976,
+      "loss": 0.6639,
+      "step": 7044
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012898222770626268,
+      "loss": 0.7508,
+      "step": 7045
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012896481794537644,
+      "loss": 0.7543,
+      "step": 7046
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012894740722611502,
+      "loss": 0.7464,
+      "step": 7047
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012892999554905447,
+      "loss": 0.757,
+      "step": 7048
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012891258291477086,
+      "loss": 0.6844,
+      "step": 7049
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012889516932384031,
+      "loss": 0.707,
+      "step": 7050
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012887775477683907,
+      "loss": 0.7194,
+      "step": 7051
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001288603392743433,
+      "loss": 0.6837,
+      "step": 7052
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012884292281692931,
+      "loss": 0.6452,
+      "step": 7053
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012882550540517325,
+      "loss": 0.6905,
+      "step": 7054
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012880808703965148,
+      "loss": 0.6301,
+      "step": 7055
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001287906677209403,
+      "loss": 0.7091,
+      "step": 7056
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012877324744961613,
+      "loss": 0.7115,
+      "step": 7057
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012875582622625534,
+      "loss": 0.7245,
+      "step": 7058
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001287384040514344,
+      "loss": 0.6939,
+      "step": 7059
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012872098092572964,
+      "loss": 0.715,
+      "step": 7060
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00012870355684971763,
+      "loss": 0.6856,
+      "step": 7061
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0001286861318239749,
+      "loss": 0.6391,
+      "step": 7062
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000128668705849078,
+      "loss": 0.7158,
+      "step": 7063
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012865127892560348,
+      "loss": 0.6814,
+      "step": 7064
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000128633851054128,
+      "loss": 0.6998,
+      "step": 7065
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001286164222352282,
+      "loss": 0.7219,
+      "step": 7066
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012859899246948065,
+      "loss": 0.694,
+      "step": 7067
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001285815617574622,
+      "loss": 0.6814,
+      "step": 7068
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012856413009974956,
+      "loss": 0.6883,
+      "step": 7069
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001285466974969195,
+      "loss": 0.7494,
+      "step": 7070
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012852926394954872,
+      "loss": 0.6886,
+      "step": 7071
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012851182945821418,
+      "loss": 0.7092,
+      "step": 7072
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012849439402349265,
+      "loss": 0.7669,
+      "step": 7073
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012847695764596113,
+      "loss": 0.6721,
+      "step": 7074
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001284595203261965,
+      "loss": 0.7304,
+      "step": 7075
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012844208206477567,
+      "loss": 0.7249,
+      "step": 7076
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001284246428622757,
+      "loss": 0.7495,
+      "step": 7077
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012840720271927353,
+      "loss": 0.7702,
+      "step": 7078
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001283897616363463,
+      "loss": 0.6669,
+      "step": 7079
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012837231961407105,
+      "loss": 0.7206,
+      "step": 7080
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012835487665302492,
+      "loss": 0.6674,
+      "step": 7081
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000128337432753785,
+      "loss": 0.7016,
+      "step": 7082
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012831998791692852,
+      "loss": 0.6798,
+      "step": 7083
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012830254214303265,
+      "loss": 0.634,
+      "step": 7084
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012828509543267468,
+      "loss": 0.7242,
+      "step": 7085
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001282676477864318,
+      "loss": 0.7133,
+      "step": 7086
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001282501992048814,
+      "loss": 0.74,
+      "step": 7087
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012823274968860074,
+      "loss": 0.6938,
+      "step": 7088
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001282152992381672,
+      "loss": 0.6987,
+      "step": 7089
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012819784785415817,
+      "loss": 0.664,
+      "step": 7090
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012818039553715114,
+      "loss": 0.7207,
+      "step": 7091
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012816294228772346,
+      "loss": 0.7155,
+      "step": 7092
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001281454881064527,
+      "loss": 0.6786,
+      "step": 7093
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012812803299391628,
+      "loss": 0.7154,
+      "step": 7094
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012811057695069183,
+      "loss": 0.7329,
+      "step": 7095
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012809311997735696,
+      "loss": 0.7778,
+      "step": 7096
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012807566207448918,
+      "loss": 0.6675,
+      "step": 7097
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012805820324266617,
+      "loss": 0.6679,
+      "step": 7098
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012804074348246562,
+      "loss": 0.6582,
+      "step": 7099
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001280232827944652,
+      "loss": 0.7499,
+      "step": 7100
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012800582117924268,
+      "loss": 0.649,
+      "step": 7101
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012798835863737576,
+      "loss": 0.7551,
+      "step": 7102
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00012797089516944233,
+      "loss": 0.673,
+      "step": 7103
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001279534307760201,
+      "loss": 0.6833,
+      "step": 7104
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.000127935965457687,
+      "loss": 0.7174,
+      "step": 7105
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012791849921502092,
+      "loss": 0.7113,
+      "step": 7106
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0001279010320485997,
+      "loss": 0.6589,
+      "step": 7107
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012788356395900137,
+      "loss": 0.6813,
+      "step": 7108
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012786609494680386,
+      "loss": 0.797,
+      "step": 7109
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012784862501258518,
+      "loss": 0.7273,
+      "step": 7110
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012783115415692338,
+      "loss": 0.7426,
+      "step": 7111
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012781368238039655,
+      "loss": 0.6628,
+      "step": 7112
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012779620968358273,
+      "loss": 0.6963,
+      "step": 7113
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012777873606706014,
+      "loss": 0.6974,
+      "step": 7114
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012776126153140683,
+      "loss": 0.7553,
+      "step": 7115
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012774378607720104,
+      "loss": 0.6456,
+      "step": 7116
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.000127726309705021,
+      "loss": 0.7243,
+      "step": 7117
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012770883241544495,
+      "loss": 0.6798,
+      "step": 7118
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012769135420905118,
+      "loss": 0.6826,
+      "step": 7119
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.000127673875086418,
+      "loss": 0.6654,
+      "step": 7120
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012765639504812374,
+      "loss": 0.6303,
+      "step": 7121
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012763891409474677,
+      "loss": 0.7419,
+      "step": 7122
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012762143222686554,
+      "loss": 0.694,
+      "step": 7123
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0001276039494450584,
+      "loss": 0.7259,
+      "step": 7124
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012758646574990385,
+      "loss": 0.6478,
+      "step": 7125
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012756898114198043,
+      "loss": 0.8025,
+      "step": 7126
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012755149562186657,
+      "loss": 0.7317,
+      "step": 7127
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012753400919014092,
+      "loss": 0.6758,
+      "step": 7128
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012751652184738197,
+      "loss": 0.7938,
+      "step": 7129
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012749903359416843,
+      "loss": 0.6596,
+      "step": 7130
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012748154443107884,
+      "loss": 0.5944,
+      "step": 7131
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012746405435869198,
+      "loss": 0.6997,
+      "step": 7132
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0001274465633775865,
+      "loss": 0.6865,
+      "step": 7133
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0001274290714883411,
+      "loss": 0.6855,
+      "step": 7134
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0001274115786915346,
+      "loss": 0.6212,
+      "step": 7135
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012739408498774576,
+      "loss": 0.6991,
+      "step": 7136
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012737659037755344,
+      "loss": 0.7263,
+      "step": 7137
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012735909486153647,
+      "loss": 0.7108,
+      "step": 7138
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012734159844027376,
+      "loss": 0.6581,
+      "step": 7139
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012732410111434414,
+      "loss": 0.7155,
+      "step": 7140
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012730660288432665,
+      "loss": 0.6847,
+      "step": 7141
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0001272891037508002,
+      "loss": 0.6751,
+      "step": 7142
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012727160371434385,
+      "loss": 0.7418,
+      "step": 7143
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012725410277553666,
+      "loss": 0.6485,
+      "step": 7144
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012723660093495753,
+      "loss": 0.6213,
+      "step": 7145
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0001272190981931857,
+      "loss": 0.7119,
+      "step": 7146
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012720159455080027,
+      "loss": 0.6758,
+      "step": 7147
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00012718409000838038,
+      "loss": 0.7881,
+      "step": 7148
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012716658456650524,
+      "loss": 0.7074,
+      "step": 7149
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012714907822575398,
+      "loss": 0.6939,
+      "step": 7150
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001271315709867059,
+      "loss": 0.7701,
+      "step": 7151
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012711406284994026,
+      "loss": 0.6953,
+      "step": 7152
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001270965538160364,
+      "loss": 0.7047,
+      "step": 7153
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001270790438855736,
+      "loss": 0.6588,
+      "step": 7154
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001270615330591313,
+      "loss": 0.7131,
+      "step": 7155
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012704402133728877,
+      "loss": 0.6618,
+      "step": 7156
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012702650872062544,
+      "loss": 0.7278,
+      "step": 7157
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012700899520972088,
+      "loss": 0.7211,
+      "step": 7158
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001269914808051545,
+      "loss": 0.786,
+      "step": 7159
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012697396550750582,
+      "loss": 0.6491,
+      "step": 7160
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012695644931735436,
+      "loss": 0.7284,
+      "step": 7161
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012693893223527966,
+      "loss": 0.778,
+      "step": 7162
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012692141426186136,
+      "loss": 0.6569,
+      "step": 7163
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001269038953976791,
+      "loss": 0.7227,
+      "step": 7164
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012688637564331256,
+      "loss": 0.6853,
+      "step": 7165
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012686885499934132,
+      "loss": 0.7614,
+      "step": 7166
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001268513334663452,
+      "loss": 0.665,
+      "step": 7167
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012683381104490384,
+      "loss": 0.6894,
+      "step": 7168
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012681628773559712,
+      "loss": 0.6797,
+      "step": 7169
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012679876353900482,
+      "loss": 0.6755,
+      "step": 7170
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012678123845570672,
+      "loss": 0.728,
+      "step": 7171
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012676371248628273,
+      "loss": 0.6711,
+      "step": 7172
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012674618563131274,
+      "loss": 0.6763,
+      "step": 7173
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012672865789137663,
+      "loss": 0.7254,
+      "step": 7174
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001267111292670544,
+      "loss": 0.7162,
+      "step": 7175
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012669359975892602,
+      "loss": 0.7406,
+      "step": 7176
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001266760693675715,
+      "loss": 0.715,
+      "step": 7177
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001266585380935708,
+      "loss": 0.6699,
+      "step": 7178
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012664100593750408,
+      "loss": 0.609,
+      "step": 7179
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001266234728999514,
+      "loss": 0.6362,
+      "step": 7180
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012660593898149297,
+      "loss": 0.6775,
+      "step": 7181
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012658840418270877,
+      "loss": 0.6595,
+      "step": 7182
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012657086850417914,
+      "loss": 0.6689,
+      "step": 7183
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012655333194648422,
+      "loss": 0.6342,
+      "step": 7184
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012653579451020423,
+      "loss": 0.6792,
+      "step": 7185
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012651825619591954,
+      "loss": 0.6527,
+      "step": 7186
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012650071700421036,
+      "loss": 0.6589,
+      "step": 7187
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012648317693565706,
+      "loss": 0.6204,
+      "step": 7188
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012646563599083996,
+      "loss": 0.6946,
+      "step": 7189
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0001264480941703395,
+      "loss": 0.6148,
+      "step": 7190
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00012643055147473604,
+      "loss": 0.661,
+      "step": 7191
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012641300790461007,
+      "loss": 0.7409,
+      "step": 7192
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012639546346054206,
+      "loss": 0.7188,
+      "step": 7193
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012637791814311248,
+      "loss": 0.7633,
+      "step": 7194
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012636037195290189,
+      "loss": 0.6584,
+      "step": 7195
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012634282489049087,
+      "loss": 0.7087,
+      "step": 7196
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012632527695645993,
+      "loss": 0.7488,
+      "step": 7197
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012630772815138978,
+      "loss": 0.7125,
+      "step": 7198
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.000126290178475861,
+      "loss": 0.6822,
+      "step": 7199
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012627262793045432,
+      "loss": 0.6466,
+      "step": 7200
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012625507651575042,
+      "loss": 0.6816,
+      "step": 7201
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012623752423233,
+      "loss": 0.6781,
+      "step": 7202
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012621997108077388,
+      "loss": 0.6586,
+      "step": 7203
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012620241706166283,
+      "loss": 0.6288,
+      "step": 7204
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012618486217557765,
+      "loss": 0.72,
+      "step": 7205
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012616730642309922,
+      "loss": 0.6867,
+      "step": 7206
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012614974980480837,
+      "loss": 0.7076,
+      "step": 7207
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012613219232128608,
+      "loss": 0.6947,
+      "step": 7208
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012611463397311322,
+      "loss": 0.7535,
+      "step": 7209
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001260970747608708,
+      "loss": 0.6187,
+      "step": 7210
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012607951468513972,
+      "loss": 0.6372,
+      "step": 7211
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001260619537465011,
+      "loss": 0.6525,
+      "step": 7212
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012604439194553596,
+      "loss": 0.6992,
+      "step": 7213
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012602682928282538,
+      "loss": 0.7385,
+      "step": 7214
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012600926575895044,
+      "loss": 0.7404,
+      "step": 7215
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012599170137449233,
+      "loss": 0.7183,
+      "step": 7216
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012597413613003213,
+      "loss": 0.657,
+      "step": 7217
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012595657002615108,
+      "loss": 0.7069,
+      "step": 7218
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001259390030634304,
+      "loss": 0.6977,
+      "step": 7219
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012592143524245132,
+      "loss": 0.7222,
+      "step": 7220
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012590386656379516,
+      "loss": 0.6582,
+      "step": 7221
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001258862970280432,
+      "loss": 0.7078,
+      "step": 7222
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001258687266357767,
+      "loss": 0.7066,
+      "step": 7223
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012585115538757714,
+      "loss": 0.6253,
+      "step": 7224
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012583358328402587,
+      "loss": 0.709,
+      "step": 7225
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012581601032570427,
+      "loss": 0.7193,
+      "step": 7226
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001257984365131938,
+      "loss": 0.6854,
+      "step": 7227
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012578086184707597,
+      "loss": 0.7671,
+      "step": 7228
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001257632863279323,
+      "loss": 0.698,
+      "step": 7229
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001257457099563442,
+      "loss": 0.7267,
+      "step": 7230
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012572813273289336,
+      "loss": 0.7074,
+      "step": 7231
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001257105546581613,
+      "loss": 0.6651,
+      "step": 7232
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00012569297573272967,
+      "loss": 0.6795,
+      "step": 7233
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001256753959571801,
+      "loss": 0.7281,
+      "step": 7234
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012565781533209423,
+      "loss": 0.7091,
+      "step": 7235
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001256402338580538,
+      "loss": 0.7435,
+      "step": 7236
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012562265153564055,
+      "loss": 0.7385,
+      "step": 7237
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012560506836543618,
+      "loss": 0.6785,
+      "step": 7238
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012558748434802252,
+      "loss": 0.6648,
+      "step": 7239
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001255698994839814,
+      "loss": 0.6739,
+      "step": 7240
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012555231377389458,
+      "loss": 0.7019,
+      "step": 7241
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.000125534727218344,
+      "loss": 0.7667,
+      "step": 7242
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012551713981791155,
+      "loss": 0.6973,
+      "step": 7243
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001254995515731791,
+      "loss": 0.6816,
+      "step": 7244
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012548196248472873,
+      "loss": 0.625,
+      "step": 7245
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012546437255314222,
+      "loss": 0.7129,
+      "step": 7246
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012544678177900176,
+      "loss": 0.7097,
+      "step": 7247
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012542919016288927,
+      "loss": 0.7261,
+      "step": 7248
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012541159770538692,
+      "loss": 0.7061,
+      "step": 7249
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012539400440707673,
+      "loss": 0.6503,
+      "step": 7250
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012537641026854078,
+      "loss": 0.6851,
+      "step": 7251
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012535881529036127,
+      "loss": 0.7314,
+      "step": 7252
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012534121947312037,
+      "loss": 0.7847,
+      "step": 7253
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012532362281740032,
+      "loss": 0.6942,
+      "step": 7254
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001253060253237833,
+      "loss": 0.7529,
+      "step": 7255
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012528842699285162,
+      "loss": 0.6709,
+      "step": 7256
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001252708278251875,
+      "loss": 0.7194,
+      "step": 7257
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012525322782137326,
+      "loss": 0.6496,
+      "step": 7258
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012523562698199127,
+      "loss": 0.7147,
+      "step": 7259
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001252180253076239,
+      "loss": 0.7165,
+      "step": 7260
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012520042279885355,
+      "loss": 0.7277,
+      "step": 7261
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012518281945626263,
+      "loss": 0.7453,
+      "step": 7262
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001251652152804336,
+      "loss": 0.6587,
+      "step": 7263
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012514761027194893,
+      "loss": 0.6761,
+      "step": 7264
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012513000443139112,
+      "loss": 0.6779,
+      "step": 7265
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012511239775934275,
+      "loss": 0.6599,
+      "step": 7266
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001250947902563863,
+      "loss": 0.7134,
+      "step": 7267
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012507718192310442,
+      "loss": 0.7023,
+      "step": 7268
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012505957276007972,
+      "loss": 0.7291,
+      "step": 7269
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0001250419627678948,
+      "loss": 0.7064,
+      "step": 7270
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012502435194713244,
+      "loss": 0.6912,
+      "step": 7271
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012500674029837524,
+      "loss": 0.7333,
+      "step": 7272
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012498912782220596,
+      "loss": 0.6731,
+      "step": 7273
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012497151451920736,
+      "loss": 0.7524,
+      "step": 7274
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012495390038996215,
+      "loss": 0.7379,
+      "step": 7275
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012493628543505322,
+      "loss": 0.7311,
+      "step": 7276
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012491866965506346,
+      "loss": 0.6986,
+      "step": 7277
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00012490105305057559,
+      "loss": 0.741,
+      "step": 7278
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012488343562217256,
+      "loss": 0.6991,
+      "step": 7279
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012486581737043736,
+      "loss": 0.7311,
+      "step": 7280
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012484819829595279,
+      "loss": 0.7081,
+      "step": 7281
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012483057839930196,
+      "loss": 0.7399,
+      "step": 7282
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001248129576810678,
+      "loss": 0.7285,
+      "step": 7283
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012479533614183334,
+      "loss": 0.7002,
+      "step": 7284
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012477771378218165,
+      "loss": 0.6674,
+      "step": 7285
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012476009060269577,
+      "loss": 0.6719,
+      "step": 7286
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001247424666039589,
+      "loss": 0.6856,
+      "step": 7287
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001247248417865541,
+      "loss": 0.7019,
+      "step": 7288
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012470721615106452,
+      "loss": 0.753,
+      "step": 7289
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012468958969807342,
+      "loss": 0.7107,
+      "step": 7290
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012467196242816396,
+      "loss": 0.7552,
+      "step": 7291
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001246543343419194,
+      "loss": 0.7759,
+      "step": 7292
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012463670543992302,
+      "loss": 0.7472,
+      "step": 7293
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012461907572275808,
+      "loss": 0.7129,
+      "step": 7294
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012460144519100793,
+      "loss": 0.7508,
+      "step": 7295
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012458381384525594,
+      "loss": 0.6747,
+      "step": 7296
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012456618168608547,
+      "loss": 0.7029,
+      "step": 7297
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012454854871407994,
+      "loss": 0.7155,
+      "step": 7298
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012453091492982274,
+      "loss": 0.6952,
+      "step": 7299
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012451328033389738,
+      "loss": 0.7139,
+      "step": 7300
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001244956449268873,
+      "loss": 0.7349,
+      "step": 7301
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012447800870937604,
+      "loss": 0.6653,
+      "step": 7302
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012446037168194714,
+      "loss": 0.715,
+      "step": 7303
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012444273384518416,
+      "loss": 0.7067,
+      "step": 7304
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012442509519967069,
+      "loss": 0.6633,
+      "step": 7305
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001244074557459904,
+      "loss": 0.7282,
+      "step": 7306
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012438981548472682,
+      "loss": 0.6728,
+      "step": 7307
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012437217441646373,
+      "loss": 0.672,
+      "step": 7308
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012435453254178478,
+      "loss": 0.6927,
+      "step": 7309
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001243368898612737,
+      "loss": 0.6965,
+      "step": 7310
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001243192463755143,
+      "loss": 0.686,
+      "step": 7311
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012430160208509028,
+      "loss": 0.7077,
+      "step": 7312
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012428395699058548,
+      "loss": 0.7177,
+      "step": 7313
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012426631109258373,
+      "loss": 0.6731,
+      "step": 7314
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001242486643916689,
+      "loss": 0.6953,
+      "step": 7315
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012423101688842486,
+      "loss": 0.697,
+      "step": 7316
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012421336858343555,
+      "loss": 0.6412,
+      "step": 7317
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012419571947728486,
+      "loss": 0.7297,
+      "step": 7318
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012417806957055685,
+      "loss": 0.658,
+      "step": 7319
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001241604188638354,
+      "loss": 0.6964,
+      "step": 7320
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00012414276735770458,
+      "loss": 0.7257,
+      "step": 7321
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012412511505274844,
+      "loss": 0.6818,
+      "step": 7322
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012410746194955105,
+      "loss": 0.6773,
+      "step": 7323
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012408980804869653,
+      "loss": 0.7349,
+      "step": 7324
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012407215335076893,
+      "loss": 0.5974,
+      "step": 7325
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012405449785635249,
+      "loss": 0.6448,
+      "step": 7326
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001240368415660313,
+      "loss": 0.7657,
+      "step": 7327
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012401918448038966,
+      "loss": 0.6741,
+      "step": 7328
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012400152660001172,
+      "loss": 0.6326,
+      "step": 7329
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012398386792548176,
+      "loss": 0.6649,
+      "step": 7330
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001239662084573841,
+      "loss": 0.6842,
+      "step": 7331
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.000123948548196303,
+      "loss": 0.6979,
+      "step": 7332
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012393088714282279,
+      "loss": 0.6864,
+      "step": 7333
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012391322529752788,
+      "loss": 0.6188,
+      "step": 7334
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012389556266100268,
+      "loss": 0.7522,
+      "step": 7335
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012387789923383146,
+      "loss": 0.7021,
+      "step": 7336
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001238602350165988,
+      "loss": 0.7379,
+      "step": 7337
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001238425700098891,
+      "loss": 0.7224,
+      "step": 7338
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012382490421428689,
+      "loss": 0.6898,
+      "step": 7339
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012380723763037667,
+      "loss": 0.6863,
+      "step": 7340
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.000123789570258743,
+      "loss": 0.6645,
+      "step": 7341
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001237719020999704,
+      "loss": 0.6167,
+      "step": 7342
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001237719020999704,
+      "loss": 0.6721,
+      "step": 7343
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012375423315464355,
+      "loss": 0.6825,
+      "step": 7344
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012373656342334698,
+      "loss": 0.7389,
+      "step": 7345
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012371889290666542,
+      "loss": 0.6574,
+      "step": 7346
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001237012216051835,
+      "loss": 0.6926,
+      "step": 7347
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012368354951948585,
+      "loss": 0.7113,
+      "step": 7348
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012366587665015738,
+      "loss": 0.6713,
+      "step": 7349
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012364820299778267,
+      "loss": 0.7057,
+      "step": 7350
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012363052856294657,
+      "loss": 0.6467,
+      "step": 7351
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012361285334623392,
+      "loss": 0.6619,
+      "step": 7352
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012359517734822947,
+      "loss": 0.702,
+      "step": 7353
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001235775005695181,
+      "loss": 0.725,
+      "step": 7354
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012355982301068472,
+      "loss": 0.6388,
+      "step": 7355
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012354214467231423,
+      "loss": 0.6762,
+      "step": 7356
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012352446555499155,
+      "loss": 0.7598,
+      "step": 7357
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012350678565930165,
+      "loss": 0.767,
+      "step": 7358
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001234891049858295,
+      "loss": 0.806,
+      "step": 7359
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0001234714235351601,
+      "loss": 0.7017,
+      "step": 7360
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012345374130787854,
+      "loss": 0.6335,
+      "step": 7361
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012343605830456982,
+      "loss": 0.7639,
+      "step": 7362
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012341837452581908,
+      "loss": 0.681,
+      "step": 7363
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00012340068997221138,
+      "loss": 0.6864,
+      "step": 7364
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012338300464433192,
+      "loss": 0.6613,
+      "step": 7365
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001233653185427658,
+      "loss": 0.6853,
+      "step": 7366
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012334763166809825,
+      "loss": 0.6718,
+      "step": 7367
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012332994402091448,
+      "loss": 0.7121,
+      "step": 7368
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012331225560179972,
+      "loss": 0.702,
+      "step": 7369
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012329456641133925,
+      "loss": 0.6966,
+      "step": 7370
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001232768764501183,
+      "loss": 0.7295,
+      "step": 7371
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012325918571872232,
+      "loss": 0.7435,
+      "step": 7372
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012324149421773654,
+      "loss": 0.6306,
+      "step": 7373
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001232238019477464,
+      "loss": 0.7149,
+      "step": 7374
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012320610890933725,
+      "loss": 0.6869,
+      "step": 7375
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001231884151030945,
+      "loss": 0.6501,
+      "step": 7376
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001231707205296036,
+      "loss": 0.6967,
+      "step": 7377
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012315302518945003,
+      "loss": 0.6186,
+      "step": 7378
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012313532908321935,
+      "loss": 0.6985,
+      "step": 7379
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.000123117632211497,
+      "loss": 0.6535,
+      "step": 7380
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001230999345748685,
+      "loss": 0.7396,
+      "step": 7381
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012308223617391953,
+      "loss": 0.6871,
+      "step": 7382
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012306453700923554,
+      "loss": 0.6971,
+      "step": 7383
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012304683708140232,
+      "loss": 0.7802,
+      "step": 7384
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012302913639100543,
+      "loss": 0.6918,
+      "step": 7385
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001230114349386305,
+      "loss": 0.7428,
+      "step": 7386
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012299373272486332,
+      "loss": 0.7033,
+      "step": 7387
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012297602975028954,
+      "loss": 0.7048,
+      "step": 7388
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012295832601549498,
+      "loss": 0.6809,
+      "step": 7389
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012294062152106534,
+      "loss": 0.7017,
+      "step": 7390
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012292291626758646,
+      "loss": 0.6572,
+      "step": 7391
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012290521025564412,
+      "loss": 0.6815,
+      "step": 7392
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012288750348582423,
+      "loss": 0.7445,
+      "step": 7393
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012286979595871263,
+      "loss": 0.7591,
+      "step": 7394
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001228520876748952,
+      "loss": 0.6593,
+      "step": 7395
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012283437863495794,
+      "loss": 0.6484,
+      "step": 7396
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012281666883948674,
+      "loss": 0.78,
+      "step": 7397
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012279895828906756,
+      "loss": 0.7549,
+      "step": 7398
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001227812469842864,
+      "loss": 0.6631,
+      "step": 7399
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012276353492572935,
+      "loss": 0.6942,
+      "step": 7400
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001227458221139824,
+      "loss": 0.7524,
+      "step": 7401
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012272810854963164,
+      "loss": 0.7678,
+      "step": 7402
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012271039423326314,
+      "loss": 0.7362,
+      "step": 7403
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012269267916546308,
+      "loss": 0.6523,
+      "step": 7404
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012267496334681757,
+      "loss": 0.7099,
+      "step": 7405
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00012265724677791278,
+      "loss": 0.7015,
+      "step": 7406
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001226395294593349,
+      "loss": 0.616,
+      "step": 7407
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012262181139167022,
+      "loss": 0.7393,
+      "step": 7408
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012260409257550489,
+      "loss": 0.7524,
+      "step": 7409
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012258637301142526,
+      "loss": 0.6903,
+      "step": 7410
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001225686527000176,
+      "loss": 0.6092,
+      "step": 7411
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012255093164186825,
+      "loss": 0.6966,
+      "step": 7412
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001225332098375635,
+      "loss": 0.645,
+      "step": 7413
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001225154872876898,
+      "loss": 0.7031,
+      "step": 7414
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012249776399283346,
+      "loss": 0.6595,
+      "step": 7415
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012248003995358098,
+      "loss": 0.8146,
+      "step": 7416
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012246231517051875,
+      "loss": 0.7125,
+      "step": 7417
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012244458964423327,
+      "loss": 0.7071,
+      "step": 7418
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012242686337531102,
+      "loss": 0.7403,
+      "step": 7419
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012240913636433852,
+      "loss": 0.7122,
+      "step": 7420
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012239140861190233,
+      "loss": 0.6478,
+      "step": 7421
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.000122373680118589,
+      "loss": 0.6854,
+      "step": 7422
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012235595088498513,
+      "loss": 0.6477,
+      "step": 7423
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012233822091167732,
+      "loss": 0.7005,
+      "step": 7424
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012232049019925226,
+      "loss": 0.7472,
+      "step": 7425
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012230275874829658,
+      "loss": 0.6894,
+      "step": 7426
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012228502655939696,
+      "loss": 0.7243,
+      "step": 7427
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012226729363314011,
+      "loss": 0.7339,
+      "step": 7428
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012224955997011281,
+      "loss": 0.7377,
+      "step": 7429
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001222318255709018,
+      "loss": 0.6598,
+      "step": 7430
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012221409043609388,
+      "loss": 0.7652,
+      "step": 7431
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012219635456627584,
+      "loss": 0.7446,
+      "step": 7432
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012217861796203454,
+      "loss": 0.6675,
+      "step": 7433
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001221608806239568,
+      "loss": 0.706,
+      "step": 7434
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012214314255262957,
+      "loss": 0.6777,
+      "step": 7435
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012212540374863972,
+      "loss": 0.7691,
+      "step": 7436
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001221076642125742,
+      "loss": 0.7246,
+      "step": 7437
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012208992394501992,
+      "loss": 0.648,
+      "step": 7438
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001220721829465639,
+      "loss": 0.7362,
+      "step": 7439
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012205444121779317,
+      "loss": 0.6325,
+      "step": 7440
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012203669875929474,
+      "loss": 0.7407,
+      "step": 7441
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012201895557165564,
+      "loss": 0.7537,
+      "step": 7442
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012200121165546299,
+      "loss": 0.6685,
+      "step": 7443
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012198346701130383,
+      "loss": 0.6649,
+      "step": 7444
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012196572163976536,
+      "loss": 0.6852,
+      "step": 7445
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001219479755414347,
+      "loss": 0.7277,
+      "step": 7446
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012193022871689904,
+      "loss": 0.717,
+      "step": 7447
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012191248116674556,
+      "loss": 0.7943,
+      "step": 7448
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012189473289156148,
+      "loss": 0.6594,
+      "step": 7449
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00012187698389193403,
+      "loss": 0.7159,
+      "step": 7450
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012185923416845053,
+      "loss": 0.718,
+      "step": 7451
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012184148372169826,
+      "loss": 0.6632,
+      "step": 7452
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012182373255226454,
+      "loss": 0.6997,
+      "step": 7453
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012180598066073667,
+      "loss": 0.66,
+      "step": 7454
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012178822804770204,
+      "loss": 0.7256,
+      "step": 7455
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012177047471374807,
+      "loss": 0.6475,
+      "step": 7456
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012175272065946217,
+      "loss": 0.7033,
+      "step": 7457
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012173496588543175,
+      "loss": 0.6268,
+      "step": 7458
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0001217172103922443,
+      "loss": 0.7518,
+      "step": 7459
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012169945418048729,
+      "loss": 0.6603,
+      "step": 7460
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0001216816972507482,
+      "loss": 0.6965,
+      "step": 7461
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012166393960361462,
+      "loss": 0.7069,
+      "step": 7462
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012164618123967409,
+      "loss": 0.7281,
+      "step": 7463
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012162842215951418,
+      "loss": 0.6702,
+      "step": 7464
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0001216106623637225,
+      "loss": 0.7089,
+      "step": 7465
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012159290185288665,
+      "loss": 0.6605,
+      "step": 7466
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012157514062759431,
+      "loss": 0.6727,
+      "step": 7467
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012155737868843318,
+      "loss": 0.7291,
+      "step": 7468
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012153961603599095,
+      "loss": 0.8059,
+      "step": 7469
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0001215218526708553,
+      "loss": 0.6849,
+      "step": 7470
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012150408859361399,
+      "loss": 0.7631,
+      "step": 7471
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0001214863238048548,
+      "loss": 0.7248,
+      "step": 7472
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012146855830516555,
+      "loss": 0.6985,
+      "step": 7473
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012145079209513402,
+      "loss": 0.7104,
+      "step": 7474
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0001214330251753481,
+      "loss": 0.6782,
+      "step": 7475
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012141525754639558,
+      "loss": 0.7157,
+      "step": 7476
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012139748920886438,
+      "loss": 0.6947,
+      "step": 7477
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012137972016334243,
+      "loss": 0.6185,
+      "step": 7478
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012136195041041765,
+      "loss": 0.7525,
+      "step": 7479
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012134417995067801,
+      "loss": 0.6294,
+      "step": 7480
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012132640878471147,
+      "loss": 0.6714,
+      "step": 7481
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012130863691310603,
+      "loss": 0.7291,
+      "step": 7482
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012129086433644976,
+      "loss": 0.6682,
+      "step": 7483
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012127309105533065,
+      "loss": 0.7352,
+      "step": 7484
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012125531707033684,
+      "loss": 0.7193,
+      "step": 7485
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012123754238205637,
+      "loss": 0.7879,
+      "step": 7486
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012121976699107743,
+      "loss": 0.6794,
+      "step": 7487
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012120199089798807,
+      "loss": 0.66,
+      "step": 7488
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012118421410337655,
+      "loss": 0.7479,
+      "step": 7489
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012116643660783098,
+      "loss": 0.7252,
+      "step": 7490
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012114865841193965,
+      "loss": 0.6558,
+      "step": 7491
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012113087951629074,
+      "loss": 0.6457,
+      "step": 7492
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012111309992147255,
+      "loss": 0.6852,
+      "step": 7493
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00012109531962807332,
+      "loss": 0.6814,
+      "step": 7494
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012107753863668139,
+      "loss": 0.6915,
+      "step": 7495
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012105975694788509,
+      "loss": 0.7656,
+      "step": 7496
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012104197456227276,
+      "loss": 0.6886,
+      "step": 7497
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012102419148043279,
+      "loss": 0.6292,
+      "step": 7498
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012100640770295355,
+      "loss": 0.6899,
+      "step": 7499
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012098862323042349,
+      "loss": 0.7533,
+      "step": 7500
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012097083806343103,
+      "loss": 0.7427,
+      "step": 7501
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012095305220256468,
+      "loss": 0.719,
+      "step": 7502
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012093526564841289,
+      "loss": 0.6344,
+      "step": 7503
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012091747840156418,
+      "loss": 0.717,
+      "step": 7504
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0001208996904626071,
+      "loss": 0.6995,
+      "step": 7505
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012088190183213022,
+      "loss": 0.7219,
+      "step": 7506
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012086411251072207,
+      "loss": 0.6465,
+      "step": 7507
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012084632249897132,
+      "loss": 0.668,
+      "step": 7508
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012082853179746657,
+      "loss": 0.7249,
+      "step": 7509
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012081074040679645,
+      "loss": 0.6825,
+      "step": 7510
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012079294832754969,
+      "loss": 0.6769,
+      "step": 7511
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012077515556031491,
+      "loss": 0.7003,
+      "step": 7512
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0001207573621056809,
+      "loss": 0.6583,
+      "step": 7513
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012073956796423635,
+      "loss": 0.7541,
+      "step": 7514
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012072177313657005,
+      "loss": 0.7148,
+      "step": 7515
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012070397762327079,
+      "loss": 0.6906,
+      "step": 7516
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012068618142492738,
+      "loss": 0.71,
+      "step": 7517
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012066838454212865,
+      "loss": 0.6283,
+      "step": 7518
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012065058697546345,
+      "loss": 0.7587,
+      "step": 7519
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012063278872552067,
+      "loss": 0.6742,
+      "step": 7520
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012061498979288916,
+      "loss": 0.6526,
+      "step": 7521
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012059719017815791,
+      "loss": 0.7222,
+      "step": 7522
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012057938988191585,
+      "loss": 0.739,
+      "step": 7523
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012056158890475194,
+      "loss": 0.7112,
+      "step": 7524
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012054378724725517,
+      "loss": 0.7168,
+      "step": 7525
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012052598491001455,
+      "loss": 0.6896,
+      "step": 7526
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012050818189361912,
+      "loss": 0.7197,
+      "step": 7527
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012049037819865792,
+      "loss": 0.7128,
+      "step": 7528
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012047257382572006,
+      "loss": 0.6697,
+      "step": 7529
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012045476877539466,
+      "loss": 0.7323,
+      "step": 7530
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012043696304827077,
+      "loss": 0.7338,
+      "step": 7531
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012041915664493761,
+      "loss": 0.7352,
+      "step": 7532
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012040134956598434,
+      "loss": 0.7213,
+      "step": 7533
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012038354181200011,
+      "loss": 0.7267,
+      "step": 7534
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0001203657333835742,
+      "loss": 0.637,
+      "step": 7535
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012034792428129578,
+      "loss": 0.7033,
+      "step": 7536
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00012033011450575416,
+      "loss": 0.7234,
+      "step": 7537
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012031230405753861,
+      "loss": 0.7052,
+      "step": 7538
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0001202944929372384,
+      "loss": 0.737,
+      "step": 7539
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0001202766811454429,
+      "loss": 0.7257,
+      "step": 7540
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012025886868274143,
+      "loss": 0.7081,
+      "step": 7541
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012024105554972341,
+      "loss": 0.7494,
+      "step": 7542
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012022324174697817,
+      "loss": 0.7436,
+      "step": 7543
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012020542727509518,
+      "loss": 0.667,
+      "step": 7544
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0001201876121346638,
+      "loss": 0.6962,
+      "step": 7545
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012016979632627357,
+      "loss": 0.7083,
+      "step": 7546
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012015197985051394,
+      "loss": 0.6735,
+      "step": 7547
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012013416270797441,
+      "loss": 0.6353,
+      "step": 7548
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012011634489924455,
+      "loss": 0.6961,
+      "step": 7549
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012009852642491383,
+      "loss": 0.6262,
+      "step": 7550
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012008070728557186,
+      "loss": 0.6675,
+      "step": 7551
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012006288748180823,
+      "loss": 0.6885,
+      "step": 7552
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012004506701421256,
+      "loss": 0.6862,
+      "step": 7553
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0001200272458833745,
+      "loss": 0.7156,
+      "step": 7554
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00012000942408988368,
+      "loss": 0.7075,
+      "step": 7555
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011999160163432974,
+      "loss": 0.7104,
+      "step": 7556
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011997377851730248,
+      "loss": 0.8032,
+      "step": 7557
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011995595473939156,
+      "loss": 0.7003,
+      "step": 7558
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011993813030118674,
+      "loss": 0.6652,
+      "step": 7559
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011992030520327779,
+      "loss": 0.6602,
+      "step": 7560
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0001199024794462545,
+      "loss": 0.6822,
+      "step": 7561
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011988465303070664,
+      "loss": 0.7403,
+      "step": 7562
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011986682595722412,
+      "loss": 0.7659,
+      "step": 7563
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011984899822639672,
+      "loss": 0.7002,
+      "step": 7564
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0001198311698388144,
+      "loss": 0.7161,
+      "step": 7565
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011981334079506696,
+      "loss": 0.625,
+      "step": 7566
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011979551109574438,
+      "loss": 0.6413,
+      "step": 7567
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011977768074143656,
+      "loss": 0.7229,
+      "step": 7568
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011975984973273353,
+      "loss": 0.6349,
+      "step": 7569
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011974201807022525,
+      "loss": 0.6893,
+      "step": 7570
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011972418575450166,
+      "loss": 0.6892,
+      "step": 7571
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011970635278615288,
+      "loss": 0.7078,
+      "step": 7572
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011968851916576888,
+      "loss": 0.6723,
+      "step": 7573
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011967068489393978,
+      "loss": 0.7066,
+      "step": 7574
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011965284997125568,
+      "loss": 0.6943,
+      "step": 7575
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011963501439830669,
+      "loss": 0.7007,
+      "step": 7576
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011961717817568292,
+      "loss": 0.6943,
+      "step": 7577
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011959934130397452,
+      "loss": 0.6712,
+      "step": 7578
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011958150378377169,
+      "loss": 0.7057,
+      "step": 7579
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00011956366561566463,
+      "loss": 0.6929,
+      "step": 7580
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011954582680024358,
+      "loss": 0.6912,
+      "step": 7581
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011952798733809875,
+      "loss": 0.6849,
+      "step": 7582
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011951014722982042,
+      "loss": 0.7083,
+      "step": 7583
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011949230647599888,
+      "loss": 0.6891,
+      "step": 7584
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011947446507722437,
+      "loss": 0.714,
+      "step": 7585
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011945662303408735,
+      "loss": 0.6835,
+      "step": 7586
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011943878034717807,
+      "loss": 0.7545,
+      "step": 7587
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011942093701708695,
+      "loss": 0.7181,
+      "step": 7588
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011940309304440433,
+      "loss": 0.7257,
+      "step": 7589
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011938524842972068,
+      "loss": 0.6653,
+      "step": 7590
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011936740317362636,
+      "loss": 0.7512,
+      "step": 7591
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011934955727671193,
+      "loss": 0.6608,
+      "step": 7592
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011933171073956781,
+      "loss": 0.7019,
+      "step": 7593
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011931386356278449,
+      "loss": 0.7572,
+      "step": 7594
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011929601574695248,
+      "loss": 0.6847,
+      "step": 7595
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011927816729266234,
+      "loss": 0.6849,
+      "step": 7596
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011926031820050468,
+      "loss": 0.6784,
+      "step": 7597
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011924246847107,
+      "loss": 0.7188,
+      "step": 7598
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011922461810494893,
+      "loss": 0.711,
+      "step": 7599
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011920676710273214,
+      "loss": 0.7657,
+      "step": 7600
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0001191889154650102,
+      "loss": 0.6745,
+      "step": 7601
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011917106319237386,
+      "loss": 0.6611,
+      "step": 7602
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011915321028541375,
+      "loss": 0.6617,
+      "step": 7603
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011913535674472059,
+      "loss": 0.7173,
+      "step": 7604
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011911750257088511,
+      "loss": 0.7139,
+      "step": 7605
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0001190996477644981,
+      "loss": 0.6724,
+      "step": 7606
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011908179232615027,
+      "loss": 0.6305,
+      "step": 7607
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011906393625643244,
+      "loss": 0.696,
+      "step": 7608
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011904607955593543,
+      "loss": 0.6795,
+      "step": 7609
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011902822222525007,
+      "loss": 0.659,
+      "step": 7610
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011901036426496723,
+      "loss": 0.7148,
+      "step": 7611
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011899250567567777,
+      "loss": 0.6202,
+      "step": 7612
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011897464645797258,
+      "loss": 0.6907,
+      "step": 7613
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011895678661244259,
+      "loss": 0.686,
+      "step": 7614
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011893892613967872,
+      "loss": 0.7012,
+      "step": 7615
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011892106504027196,
+      "loss": 0.6816,
+      "step": 7616
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011890320331481328,
+      "loss": 0.7496,
+      "step": 7617
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011888534096389367,
+      "loss": 0.6333,
+      "step": 7618
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011886747798810419,
+      "loss": 0.6314,
+      "step": 7619
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011884961438803581,
+      "loss": 0.6837,
+      "step": 7620
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011883175016427965,
+      "loss": 0.6795,
+      "step": 7621
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011881388531742679,
+      "loss": 0.6788,
+      "step": 7622
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00011879601984806829,
+      "loss": 0.6636,
+      "step": 7623
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011877815375679535,
+      "loss": 0.7152,
+      "step": 7624
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011876028704419903,
+      "loss": 0.6818,
+      "step": 7625
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011874241971087057,
+      "loss": 0.6278,
+      "step": 7626
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011872455175740112,
+      "loss": 0.7818,
+      "step": 7627
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011870668318438187,
+      "loss": 0.7117,
+      "step": 7628
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011868881399240408,
+      "loss": 0.7327,
+      "step": 7629
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.000118670944182059,
+      "loss": 0.6879,
+      "step": 7630
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011865307375393788,
+      "loss": 0.6786,
+      "step": 7631
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.000118635202708632,
+      "loss": 0.7038,
+      "step": 7632
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011861733104673269,
+      "loss": 0.7041,
+      "step": 7633
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011859945876883132,
+      "loss": 0.6297,
+      "step": 7634
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011858158587551913,
+      "loss": 0.6628,
+      "step": 7635
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011856371236738757,
+      "loss": 0.7257,
+      "step": 7636
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011854583824502803,
+      "loss": 0.6992,
+      "step": 7637
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011852796350903188,
+      "loss": 0.7528,
+      "step": 7638
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011851008815999064,
+      "loss": 0.6716,
+      "step": 7639
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011849221219849562,
+      "loss": 0.6645,
+      "step": 7640
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011847433562513842,
+      "loss": 0.7471,
+      "step": 7641
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011845645844051045,
+      "loss": 0.7143,
+      "step": 7642
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011843858064520329,
+      "loss": 0.7137,
+      "step": 7643
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011842070223980841,
+      "loss": 0.729,
+      "step": 7644
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011840282322491744,
+      "loss": 0.6139,
+      "step": 7645
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011838494360112185,
+      "loss": 0.6664,
+      "step": 7646
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011836706336901331,
+      "loss": 0.734,
+      "step": 7647
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0001183491825291834,
+      "loss": 0.7029,
+      "step": 7648
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011833130108222378,
+      "loss": 0.7023,
+      "step": 7649
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0001183134190287261,
+      "loss": 0.6773,
+      "step": 7650
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011829553636928202,
+      "loss": 0.7142,
+      "step": 7651
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0001182776531044832,
+      "loss": 0.7583,
+      "step": 7652
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011825976923492141,
+      "loss": 0.7222,
+      "step": 7653
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011824188476118835,
+      "loss": 0.6929,
+      "step": 7654
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011822399968387582,
+      "loss": 0.765,
+      "step": 7655
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011820611400357555,
+      "loss": 0.6642,
+      "step": 7656
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011818822772087932,
+      "loss": 0.6793,
+      "step": 7657
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011817034083637897,
+      "loss": 0.6872,
+      "step": 7658
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011815245335066634,
+      "loss": 0.656,
+      "step": 7659
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011813456526433327,
+      "loss": 0.6572,
+      "step": 7660
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011811667657797166,
+      "loss": 0.8045,
+      "step": 7661
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011809878729217337,
+      "loss": 0.6079,
+      "step": 7662
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011808089740753028,
+      "loss": 0.6959,
+      "step": 7663
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0001180630069246344,
+      "loss": 0.818,
+      "step": 7664
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00011804511584407763,
+      "loss": 0.6944,
+      "step": 7665
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.000118027224166452,
+      "loss": 0.6983,
+      "step": 7666
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011800933189234943,
+      "loss": 0.6841,
+      "step": 7667
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011799143902236198,
+      "loss": 0.6799,
+      "step": 7668
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011797354555708164,
+      "loss": 0.6929,
+      "step": 7669
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011795565149710051,
+      "loss": 0.6787,
+      "step": 7670
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011793775684301063,
+      "loss": 0.6267,
+      "step": 7671
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0001179198615954041,
+      "loss": 0.6971,
+      "step": 7672
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011790196575487303,
+      "loss": 0.6223,
+      "step": 7673
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011788406932200954,
+      "loss": 0.692,
+      "step": 7674
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011786617229740578,
+      "loss": 0.7,
+      "step": 7675
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011784827468165394,
+      "loss": 0.6503,
+      "step": 7676
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0001178303764753462,
+      "loss": 0.6988,
+      "step": 7677
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011781247767907475,
+      "loss": 0.7673,
+      "step": 7678
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011779457829343184,
+      "loss": 0.7266,
+      "step": 7679
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011777667831900966,
+      "loss": 0.7317,
+      "step": 7680
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011775877775640056,
+      "loss": 0.6745,
+      "step": 7681
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011774087660619683,
+      "loss": 0.7039,
+      "step": 7682
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011772297486899069,
+      "loss": 0.7143,
+      "step": 7683
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011770507254537453,
+      "loss": 0.7768,
+      "step": 7684
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011768716963594064,
+      "loss": 0.6864,
+      "step": 7685
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011766926614128142,
+      "loss": 0.7074,
+      "step": 7686
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0001176513620619893,
+      "loss": 0.645,
+      "step": 7687
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0001176334573986566,
+      "loss": 0.6786,
+      "step": 7688
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011761555215187578,
+      "loss": 0.6776,
+      "step": 7689
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011759764632223926,
+      "loss": 0.6921,
+      "step": 7690
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011757973991033955,
+      "loss": 0.7796,
+      "step": 7691
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011756183291676904,
+      "loss": 0.5879,
+      "step": 7692
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011754392534212033,
+      "loss": 0.7188,
+      "step": 7693
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011752601718698588,
+      "loss": 0.7347,
+      "step": 7694
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011750810845195826,
+      "loss": 0.7456,
+      "step": 7695
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011749019913762997,
+      "loss": 0.6796,
+      "step": 7696
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011747228924459361,
+      "loss": 0.6655,
+      "step": 7697
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011745437877344186,
+      "loss": 0.639,
+      "step": 7698
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011743646772476719,
+      "loss": 0.8453,
+      "step": 7699
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011741855609916233,
+      "loss": 0.7034,
+      "step": 7700
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011740064389721989,
+      "loss": 0.676,
+      "step": 7701
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011738273111953257,
+      "loss": 0.7012,
+      "step": 7702
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011736481776669306,
+      "loss": 0.6998,
+      "step": 7703
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011734690383929403,
+      "loss": 0.6731,
+      "step": 7704
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011732898933792824,
+      "loss": 0.6968,
+      "step": 7705
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011731107426318844,
+      "loss": 0.6554,
+      "step": 7706
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011729315861566737,
+      "loss": 0.7303,
+      "step": 7707
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011727524239595784,
+      "loss": 0.7413,
+      "step": 7708
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00011725732560465263,
+      "loss": 0.698,
+      "step": 7709
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0001172394082423446,
+      "loss": 0.6967,
+      "step": 7710
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011722149030962657,
+      "loss": 0.6847,
+      "step": 7711
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0001172035718070914,
+      "loss": 0.6499,
+      "step": 7712
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011718565273533197,
+      "loss": 0.7373,
+      "step": 7713
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011716773309494118,
+      "loss": 0.6831,
+      "step": 7714
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011714981288651194,
+      "loss": 0.6783,
+      "step": 7715
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0001171318921106372,
+      "loss": 0.6303,
+      "step": 7716
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011711397076790991,
+      "loss": 0.7064,
+      "step": 7717
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011709604885892303,
+      "loss": 0.6247,
+      "step": 7718
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011707812638426957,
+      "loss": 0.6922,
+      "step": 7719
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011706020334454253,
+      "loss": 0.6592,
+      "step": 7720
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011704227974033495,
+      "loss": 0.6865,
+      "step": 7721
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011702435557223987,
+      "loss": 0.6719,
+      "step": 7722
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011700643084085037,
+      "loss": 0.7296,
+      "step": 7723
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011698850554675951,
+      "loss": 0.7008,
+      "step": 7724
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0001169705796905604,
+      "loss": 0.6491,
+      "step": 7725
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011695265327284619,
+      "loss": 0.768,
+      "step": 7726
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011693472629421001,
+      "loss": 0.6943,
+      "step": 7727
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011691679875524499,
+      "loss": 0.7362,
+      "step": 7728
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011689887065654434,
+      "loss": 0.7125,
+      "step": 7729
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011688094199870127,
+      "loss": 0.6521,
+      "step": 7730
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011686301278230895,
+      "loss": 0.7146,
+      "step": 7731
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011684508300796066,
+      "loss": 0.7028,
+      "step": 7732
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011682715267624962,
+      "loss": 0.7456,
+      "step": 7733
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011680922178776911,
+      "loss": 0.631,
+      "step": 7734
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011679129034311244,
+      "loss": 0.686,
+      "step": 7735
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011677335834287286,
+      "loss": 0.6973,
+      "step": 7736
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011675542578764378,
+      "loss": 0.7038,
+      "step": 7737
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011673749267801848,
+      "loss": 0.7178,
+      "step": 7738
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011671955901459035,
+      "loss": 0.7205,
+      "step": 7739
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011670162479795277,
+      "loss": 0.6436,
+      "step": 7740
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011668369002869912,
+      "loss": 0.6639,
+      "step": 7741
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011666575470742283,
+      "loss": 0.6981,
+      "step": 7742
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011664781883471734,
+      "loss": 0.7171,
+      "step": 7743
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011662988241117612,
+      "loss": 0.7651,
+      "step": 7744
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0001166119454373926,
+      "loss": 0.6655,
+      "step": 7745
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011659400791396033,
+      "loss": 0.7636,
+      "step": 7746
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011657606984147274,
+      "loss": 0.6727,
+      "step": 7747
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011655813122052341,
+      "loss": 0.6783,
+      "step": 7748
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0001165401920517059,
+      "loss": 0.6918,
+      "step": 7749
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011652225233561374,
+      "loss": 0.7316,
+      "step": 7750
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011650431207284053,
+      "loss": 0.6967,
+      "step": 7751
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011648637126397988,
+      "loss": 0.6396,
+      "step": 7752
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00011646842990962534,
+      "loss": 0.6997,
+      "step": 7753
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001164504880103706,
+      "loss": 0.7093,
+      "step": 7754
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011643254556680934,
+      "loss": 0.6436,
+      "step": 7755
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001164146025795352,
+      "loss": 0.6582,
+      "step": 7756
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011639665904914185,
+      "loss": 0.6544,
+      "step": 7757
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011637871497622303,
+      "loss": 0.6915,
+      "step": 7758
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011636077036137243,
+      "loss": 0.6955,
+      "step": 7759
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011634282520518383,
+      "loss": 0.6632,
+      "step": 7760
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011632487950825097,
+      "loss": 0.7227,
+      "step": 7761
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011630693327116768,
+      "loss": 0.6747,
+      "step": 7762
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011628898649452769,
+      "loss": 0.7135,
+      "step": 7763
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011627103917892479,
+      "loss": 0.7354,
+      "step": 7764
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011625309132495292,
+      "loss": 0.7318,
+      "step": 7765
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011623514293320585,
+      "loss": 0.7313,
+      "step": 7766
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001162171940042775,
+      "loss": 0.6599,
+      "step": 7767
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011619924453876172,
+      "loss": 0.7367,
+      "step": 7768
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011618129453725242,
+      "loss": 0.7315,
+      "step": 7769
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011616334400034348,
+      "loss": 0.7517,
+      "step": 7770
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011614539292862892,
+      "loss": 0.7187,
+      "step": 7771
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011612744132270267,
+      "loss": 0.721,
+      "step": 7772
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001161094891831587,
+      "loss": 0.7033,
+      "step": 7773
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011609153651059097,
+      "loss": 0.6384,
+      "step": 7774
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011607358330559355,
+      "loss": 0.7309,
+      "step": 7775
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011605562956876039,
+      "loss": 0.7204,
+      "step": 7776
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011603767530068561,
+      "loss": 0.6359,
+      "step": 7777
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011601972050196328,
+      "loss": 0.7002,
+      "step": 7778
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011600176517318741,
+      "loss": 0.7541,
+      "step": 7779
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011598380931495212,
+      "loss": 0.7869,
+      "step": 7780
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011596585292785153,
+      "loss": 0.6959,
+      "step": 7781
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011594789601247982,
+      "loss": 0.6999,
+      "step": 7782
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011592993856943111,
+      "loss": 0.7281,
+      "step": 7783
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011591198059929954,
+      "loss": 0.6686,
+      "step": 7784
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001158940221026793,
+      "loss": 0.6332,
+      "step": 7785
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011587606308016463,
+      "loss": 0.7178,
+      "step": 7786
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011585810353234969,
+      "loss": 0.6904,
+      "step": 7787
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001158401434598288,
+      "loss": 0.6355,
+      "step": 7788
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011582218286319619,
+      "loss": 0.7485,
+      "step": 7789
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011580422174304608,
+      "loss": 0.6629,
+      "step": 7790
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001157862600999728,
+      "loss": 0.6321,
+      "step": 7791
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011576829793457064,
+      "loss": 0.7074,
+      "step": 7792
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011575033524743393,
+      "loss": 0.6711,
+      "step": 7793
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011573237203915706,
+      "loss": 0.7039,
+      "step": 7794
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00011571440831033431,
+      "loss": 0.6886,
+      "step": 7795
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0001156964440615601,
+      "loss": 0.6379,
+      "step": 7796
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011567847929342881,
+      "loss": 0.7131,
+      "step": 7797
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011566051400653486,
+      "loss": 0.7066,
+      "step": 7798
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011564254820147266,
+      "loss": 0.6734,
+      "step": 7799
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011562458187883669,
+      "loss": 0.6896,
+      "step": 7800
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011560661503922139,
+      "loss": 0.6604,
+      "step": 7801
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011558864768322122,
+      "loss": 0.7185,
+      "step": 7802
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001155706798114307,
+      "loss": 0.7344,
+      "step": 7803
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011555271142444433,
+      "loss": 0.6675,
+      "step": 7804
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011553474252285666,
+      "loss": 0.7019,
+      "step": 7805
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001155167731072622,
+      "loss": 0.6483,
+      "step": 7806
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011549880317825556,
+      "loss": 0.6972,
+      "step": 7807
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011548083273643131,
+      "loss": 0.6415,
+      "step": 7808
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011546286178238403,
+      "loss": 0.6598,
+      "step": 7809
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011546286178238403,
+      "loss": 0.7156,
+      "step": 7810
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011544489031670833,
+      "loss": 0.6952,
+      "step": 7811
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011542691833999886,
+      "loss": 0.6969,
+      "step": 7812
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011540894585285027,
+      "loss": 0.6834,
+      "step": 7813
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011539097285585723,
+      "loss": 0.7135,
+      "step": 7814
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011537299934961438,
+      "loss": 0.695,
+      "step": 7815
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011535502533471647,
+      "loss": 0.674,
+      "step": 7816
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011533705081175821,
+      "loss": 0.6386,
+      "step": 7817
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011531907578133429,
+      "loss": 0.6756,
+      "step": 7818
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011530110024403952,
+      "loss": 0.6662,
+      "step": 7819
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011528312420046862,
+      "loss": 0.6895,
+      "step": 7820
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011526514765121641,
+      "loss": 0.7407,
+      "step": 7821
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011524717059687766,
+      "loss": 0.6744,
+      "step": 7822
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001152291930380472,
+      "loss": 0.683,
+      "step": 7823
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011521121497531989,
+      "loss": 0.7293,
+      "step": 7824
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011519323640929052,
+      "loss": 0.694,
+      "step": 7825
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011517525734055398,
+      "loss": 0.612,
+      "step": 7826
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011515727776970517,
+      "loss": 0.6578,
+      "step": 7827
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.000115139297697339,
+      "loss": 0.6328,
+      "step": 7828
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011512131712405037,
+      "loss": 0.6578,
+      "step": 7829
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001151033360504342,
+      "loss": 0.6599,
+      "step": 7830
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011508535447708546,
+      "loss": 0.7038,
+      "step": 7831
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001150673724045991,
+      "loss": 0.7387,
+      "step": 7832
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011504938983357012,
+      "loss": 0.7388,
+      "step": 7833
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001150314067645935,
+      "loss": 0.7252,
+      "step": 7834
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011501342319826429,
+      "loss": 0.7216,
+      "step": 7835
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0001149954391351775,
+      "loss": 0.7271,
+      "step": 7836
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011497745457592816,
+      "loss": 0.6767,
+      "step": 7837
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011495946952111136,
+      "loss": 0.6949,
+      "step": 7838
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00011494148397132219,
+      "loss": 0.7096,
+      "step": 7839
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011492349792715571,
+      "loss": 0.7542,
+      "step": 7840
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011490551138920706,
+      "loss": 0.6362,
+      "step": 7841
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011488752435807142,
+      "loss": 0.6168,
+      "step": 7842
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011486953683434382,
+      "loss": 0.6883,
+      "step": 7843
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0001148515488186195,
+      "loss": 0.688,
+      "step": 7844
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011483356031149366,
+      "loss": 0.6814,
+      "step": 7845
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011481557131356145,
+      "loss": 0.7788,
+      "step": 7846
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011479758182541809,
+      "loss": 0.7592,
+      "step": 7847
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011477959184765883,
+      "loss": 0.7502,
+      "step": 7848
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011476160138087889,
+      "loss": 0.6671,
+      "step": 7849
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011474361042567352,
+      "loss": 0.7419,
+      "step": 7850
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011472561898263805,
+      "loss": 0.7018,
+      "step": 7851
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011470762705236772,
+      "loss": 0.72,
+      "step": 7852
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0001146896346354579,
+      "loss": 0.6757,
+      "step": 7853
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011467164173250385,
+      "loss": 0.7096,
+      "step": 7854
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011465364834410089,
+      "loss": 0.6537,
+      "step": 7855
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011463565447084445,
+      "loss": 0.6959,
+      "step": 7856
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0001146176601133299,
+      "loss": 0.6752,
+      "step": 7857
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011459966527215263,
+      "loss": 0.6969,
+      "step": 7858
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011458166994790797,
+      "loss": 0.7338,
+      "step": 7859
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0001145636741411914,
+      "loss": 0.7537,
+      "step": 7860
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011454567785259836,
+      "loss": 0.675,
+      "step": 7861
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011452768108272427,
+      "loss": 0.651,
+      "step": 7862
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011450968383216465,
+      "loss": 0.6974,
+      "step": 7863
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011449168610151497,
+      "loss": 0.6963,
+      "step": 7864
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011447368789137072,
+      "loss": 0.7085,
+      "step": 7865
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011445568920232737,
+      "loss": 0.698,
+      "step": 7866
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011443769003498053,
+      "loss": 0.6615,
+      "step": 7867
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0001144196903899257,
+      "loss": 0.6748,
+      "step": 7868
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011440169026775851,
+      "loss": 0.6864,
+      "step": 7869
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011438368966907445,
+      "loss": 0.6554,
+      "step": 7870
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011436568859446917,
+      "loss": 0.6723,
+      "step": 7871
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011434768704453824,
+      "loss": 0.6695,
+      "step": 7872
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011432968501987735,
+      "loss": 0.7338,
+      "step": 7873
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0001143116825210821,
+      "loss": 0.7373,
+      "step": 7874
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011429367954874819,
+      "loss": 0.6464,
+      "step": 7875
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011427567610347124,
+      "loss": 0.6657,
+      "step": 7876
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011425767218584696,
+      "loss": 0.7188,
+      "step": 7877
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011423966779647102,
+      "loss": 0.7334,
+      "step": 7878
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0001142216629359392,
+      "loss": 0.7222,
+      "step": 7879
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011420365760484728,
+      "loss": 0.7057,
+      "step": 7880
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0001141856518037909,
+      "loss": 0.6349,
+      "step": 7881
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00011416764553336588,
+      "loss": 0.7533,
+      "step": 7882
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.000114149638794168,
+      "loss": 0.691,
+      "step": 7883
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011413163158679305,
+      "loss": 0.6457,
+      "step": 7884
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011411362391183689,
+      "loss": 0.6974,
+      "step": 7885
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011409561576989529,
+      "loss": 0.6153,
+      "step": 7886
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011407760716156412,
+      "loss": 0.6299,
+      "step": 7887
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011405959808743926,
+      "loss": 0.6277,
+      "step": 7888
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011404158854811653,
+      "loss": 0.6306,
+      "step": 7889
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011402357854419193,
+      "loss": 0.7213,
+      "step": 7890
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011400556807626123,
+      "loss": 0.7057,
+      "step": 7891
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011398755714492044,
+      "loss": 0.7414,
+      "step": 7892
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011396954575076551,
+      "loss": 0.7095,
+      "step": 7893
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011395153389439233,
+      "loss": 0.6753,
+      "step": 7894
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011393352157639687,
+      "loss": 0.7281,
+      "step": 7895
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0001139155087973752,
+      "loss": 0.7338,
+      "step": 7896
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011389749555792324,
+      "loss": 0.7149,
+      "step": 7897
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011387948185863701,
+      "loss": 0.6699,
+      "step": 7898
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011386146770011257,
+      "loss": 0.748,
+      "step": 7899
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011384345308294595,
+      "loss": 0.6411,
+      "step": 7900
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0001138254380077332,
+      "loss": 0.6263,
+      "step": 7901
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011380742247507042,
+      "loss": 0.7095,
+      "step": 7902
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011378940648555368,
+      "loss": 0.7065,
+      "step": 7903
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011377139003977909,
+      "loss": 0.751,
+      "step": 7904
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011375337313834275,
+      "loss": 0.6513,
+      "step": 7905
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011373535578184082,
+      "loss": 0.7342,
+      "step": 7906
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011371733797086948,
+      "loss": 0.6911,
+      "step": 7907
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011369931970602483,
+      "loss": 0.6741,
+      "step": 7908
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011368130098790308,
+      "loss": 0.6671,
+      "step": 7909
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011366328181710044,
+      "loss": 0.6577,
+      "step": 7910
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011364526219421309,
+      "loss": 0.6549,
+      "step": 7911
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011362724211983731,
+      "loss": 0.6594,
+      "step": 7912
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011360922159456928,
+      "loss": 0.6648,
+      "step": 7913
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011359120061900527,
+      "loss": 0.7032,
+      "step": 7914
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011357317919374157,
+      "loss": 0.6511,
+      "step": 7915
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011355515731937445,
+      "loss": 0.6806,
+      "step": 7916
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011353713499650023,
+      "loss": 0.7213,
+      "step": 7917
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011351911222571518,
+      "loss": 0.6776,
+      "step": 7918
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011350108900761569,
+      "loss": 0.6635,
+      "step": 7919
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011348306534279805,
+      "loss": 0.751,
+      "step": 7920
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011346504123185866,
+      "loss": 0.7389,
+      "step": 7921
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011344701667539385,
+      "loss": 0.7146,
+      "step": 7922
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011342899167400005,
+      "loss": 0.7699,
+      "step": 7923
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011341096622827365,
+      "loss": 0.7455,
+      "step": 7924
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011339294033881105,
+      "loss": 0.657,
+      "step": 7925
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00011337491400620872,
+      "loss": 0.7049,
+      "step": 7926
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011335688723106306,
+      "loss": 0.6903,
+      "step": 7927
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011333886001397058,
+      "loss": 0.7545,
+      "step": 7928
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011332083235552773,
+      "loss": 0.7212,
+      "step": 7929
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.000113302804256331,
+      "loss": 0.7268,
+      "step": 7930
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011328477571697691,
+      "loss": 0.7572,
+      "step": 7931
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011326674673806195,
+      "loss": 0.7369,
+      "step": 7932
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0001132487173201827,
+      "loss": 0.6515,
+      "step": 7933
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011323068746393568,
+      "loss": 0.6623,
+      "step": 7934
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011321265716991744,
+      "loss": 0.701,
+      "step": 7935
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011319462643872461,
+      "loss": 0.7228,
+      "step": 7936
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011317659527095375,
+      "loss": 0.6903,
+      "step": 7937
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011315856366720147,
+      "loss": 0.5927,
+      "step": 7938
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011314053162806436,
+      "loss": 0.6781,
+      "step": 7939
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0001131224991541391,
+      "loss": 0.7125,
+      "step": 7940
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011310446624602233,
+      "loss": 0.6706,
+      "step": 7941
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011308643290431074,
+      "loss": 0.6955,
+      "step": 7942
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011306839912960096,
+      "loss": 0.6442,
+      "step": 7943
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011305036492248972,
+      "loss": 0.6938,
+      "step": 7944
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011303233028357367,
+      "loss": 0.7725,
+      "step": 7945
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011301429521344962,
+      "loss": 0.6808,
+      "step": 7946
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011299625971271424,
+      "loss": 0.7503,
+      "step": 7947
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011297822378196431,
+      "loss": 0.6804,
+      "step": 7948
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011296018742179662,
+      "loss": 0.729,
+      "step": 7949
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011294215063280787,
+      "loss": 0.6756,
+      "step": 7950
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0001129241134155949,
+      "loss": 0.6743,
+      "step": 7951
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011290607577075453,
+      "loss": 0.6854,
+      "step": 7952
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011288803769888359,
+      "loss": 0.7063,
+      "step": 7953
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0001128699992005789,
+      "loss": 0.648,
+      "step": 7954
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0001128519602764373,
+      "loss": 0.645,
+      "step": 7955
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011283392092705562,
+      "loss": 0.6509,
+      "step": 7956
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011281588115303081,
+      "loss": 0.6519,
+      "step": 7957
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011279784095495973,
+      "loss": 0.6709,
+      "step": 7958
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011277980033343932,
+      "loss": 0.6868,
+      "step": 7959
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011276175928906641,
+      "loss": 0.6717,
+      "step": 7960
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011274371782243802,
+      "loss": 0.6518,
+      "step": 7961
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011272567593415105,
+      "loss": 0.6477,
+      "step": 7962
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011270763362480248,
+      "loss": 0.6683,
+      "step": 7963
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011268959089498929,
+      "loss": 0.6861,
+      "step": 7964
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011267154774530851,
+      "loss": 0.7344,
+      "step": 7965
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011265350417635706,
+      "loss": 0.709,
+      "step": 7966
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.000112635460188732,
+      "loss": 0.7061,
+      "step": 7967
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011261741578303033,
+      "loss": 0.7484,
+      "step": 7968
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00011259937095984918,
+      "loss": 0.7388,
+      "step": 7969
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011258132571978555,
+      "loss": 0.7097,
+      "step": 7970
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001125632800634365,
+      "loss": 0.6863,
+      "step": 7971
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011254523399139913,
+      "loss": 0.6253,
+      "step": 7972
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011252718750427053,
+      "loss": 0.7508,
+      "step": 7973
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011250914060264785,
+      "loss": 0.6856,
+      "step": 7974
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011249109328712823,
+      "loss": 0.6813,
+      "step": 7975
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011247304555830876,
+      "loss": 0.7235,
+      "step": 7976
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001124549974167866,
+      "loss": 0.6372,
+      "step": 7977
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011243694886315894,
+      "loss": 0.7022,
+      "step": 7978
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011241889989802293,
+      "loss": 0.6886,
+      "step": 7979
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001124008505219758,
+      "loss": 0.6607,
+      "step": 7980
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001123828007356148,
+      "loss": 0.7181,
+      "step": 7981
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011236475053953708,
+      "loss": 0.6705,
+      "step": 7982
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011234669993433988,
+      "loss": 0.7424,
+      "step": 7983
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001123286489206205,
+      "loss": 0.6838,
+      "step": 7984
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011231059749897615,
+      "loss": 0.7195,
+      "step": 7985
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011229254567000415,
+      "loss": 0.6858,
+      "step": 7986
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011227449343430179,
+      "loss": 0.6793,
+      "step": 7987
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011225644079246633,
+      "loss": 0.6543,
+      "step": 7988
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011223838774509514,
+      "loss": 0.6617,
+      "step": 7989
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011222033429278552,
+      "loss": 0.6481,
+      "step": 7990
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011220228043613484,
+      "loss": 0.7147,
+      "step": 7991
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011218422617574043,
+      "loss": 0.6144,
+      "step": 7992
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011216617151219968,
+      "loss": 0.7352,
+      "step": 7993
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011214811644610994,
+      "loss": 0.7167,
+      "step": 7994
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011213006097806866,
+      "loss": 0.7488,
+      "step": 7995
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011211200510867321,
+      "loss": 0.6556,
+      "step": 7996
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011209394883852108,
+      "loss": 0.8032,
+      "step": 7997
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011207589216820964,
+      "loss": 0.6996,
+      "step": 7998
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011205783509833634,
+      "loss": 0.6421,
+      "step": 7999
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011203977762949869,
+      "loss": 0.6838,
+      "step": 8000
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011202171976229412,
+      "loss": 0.7191,
+      "step": 8001
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011200366149732015,
+      "loss": 0.7348,
+      "step": 8002
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001119856028351743,
+      "loss": 0.6394,
+      "step": 8003
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011196754377645406,
+      "loss": 0.6458,
+      "step": 8004
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011194948432175696,
+      "loss": 0.6823,
+      "step": 8005
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011193142447168055,
+      "loss": 0.6782,
+      "step": 8006
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011191336422682237,
+      "loss": 0.6861,
+      "step": 8007
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011189530358778005,
+      "loss": 0.7068,
+      "step": 8008
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011187724255515108,
+      "loss": 0.7278,
+      "step": 8009
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011185918112953313,
+      "loss": 0.6315,
+      "step": 8010
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011184111931152377,
+      "loss": 0.6444,
+      "step": 8011
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00011182305710172063,
+      "loss": 0.75,
+      "step": 8012
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011180499450072137,
+      "loss": 0.7208,
+      "step": 8013
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011178693150912361,
+      "loss": 0.6716,
+      "step": 8014
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011176886812752499,
+      "loss": 0.6461,
+      "step": 8015
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011175080435652323,
+      "loss": 0.7052,
+      "step": 8016
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011173274019671601,
+      "loss": 0.7156,
+      "step": 8017
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011171467564870098,
+      "loss": 0.655,
+      "step": 8018
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0001116966107130759,
+      "loss": 0.6815,
+      "step": 8019
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0001116785453904385,
+      "loss": 0.7312,
+      "step": 8020
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011166047968138646,
+      "loss": 0.6449,
+      "step": 8021
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011164241358651759,
+      "loss": 0.6354,
+      "step": 8022
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011162434710642963,
+      "loss": 0.6635,
+      "step": 8023
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011160628024172037,
+      "loss": 0.6748,
+      "step": 8024
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011158821299298759,
+      "loss": 0.6417,
+      "step": 8025
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011157014536082906,
+      "loss": 0.6854,
+      "step": 8026
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011155207734584263,
+      "loss": 0.671,
+      "step": 8027
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011153400894862613,
+      "loss": 0.6521,
+      "step": 8028
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011151594016977738,
+      "loss": 0.6795,
+      "step": 8029
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011149787100989424,
+      "loss": 0.6798,
+      "step": 8030
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011147980146957458,
+      "loss": 0.7111,
+      "step": 8031
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011146173154941625,
+      "loss": 0.7107,
+      "step": 8032
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011144366125001717,
+      "loss": 0.6777,
+      "step": 8033
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011142559057197523,
+      "loss": 0.7204,
+      "step": 8034
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011140751951588837,
+      "loss": 0.654,
+      "step": 8035
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011138944808235448,
+      "loss": 0.6246,
+      "step": 8036
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0001113713762719715,
+      "loss": 0.677,
+      "step": 8037
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0001113533040853374,
+      "loss": 0.7267,
+      "step": 8038
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011133523152305017,
+      "loss": 0.7078,
+      "step": 8039
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011131715858570771,
+      "loss": 0.7073,
+      "step": 8040
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011129908527390809,
+      "loss": 0.72,
+      "step": 8041
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011128101158824926,
+      "loss": 0.747,
+      "step": 8042
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011126293752932926,
+      "loss": 0.7441,
+      "step": 8043
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011124486309774611,
+      "loss": 0.7443,
+      "step": 8044
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011122678829409787,
+      "loss": 0.7056,
+      "step": 8045
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011120871311898254,
+      "loss": 0.6835,
+      "step": 8046
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011119063757299822,
+      "loss": 0.7206,
+      "step": 8047
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011117256165674299,
+      "loss": 0.7305,
+      "step": 8048
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0001111544853708149,
+      "loss": 0.7182,
+      "step": 8049
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011113640871581212,
+      "loss": 0.6087,
+      "step": 8050
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0001111183316923327,
+      "loss": 0.7595,
+      "step": 8051
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011110025430097474,
+      "loss": 0.7166,
+      "step": 8052
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011108217654233649,
+      "loss": 0.7172,
+      "step": 8053
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.000111064098417016,
+      "loss": 0.6104,
+      "step": 8054
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00011104601992561147,
+      "loss": 0.7042,
+      "step": 8055
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011102794106872106,
+      "loss": 0.7304,
+      "step": 8056
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011100986184694293,
+      "loss": 0.6527,
+      "step": 8057
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011099178226087535,
+      "loss": 0.6952,
+      "step": 8058
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011097370231111646,
+      "loss": 0.6712,
+      "step": 8059
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011095562199826452,
+      "loss": 0.6508,
+      "step": 8060
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011093754132291776,
+      "loss": 0.61,
+      "step": 8061
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011091946028567442,
+      "loss": 0.6699,
+      "step": 8062
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001109013788871327,
+      "loss": 0.6657,
+      "step": 8063
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011088329712789096,
+      "loss": 0.6796,
+      "step": 8064
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011086521500854745,
+      "loss": 0.6725,
+      "step": 8065
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011084713252970049,
+      "loss": 0.6317,
+      "step": 8066
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011082904969194832,
+      "loss": 0.6677,
+      "step": 8067
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001108109664958893,
+      "loss": 0.6819,
+      "step": 8068
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011079288294212171,
+      "loss": 0.6325,
+      "step": 8069
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011077479903124394,
+      "loss": 0.7208,
+      "step": 8070
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011075671476385438,
+      "loss": 0.6648,
+      "step": 8071
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011073863014055132,
+      "loss": 0.7212,
+      "step": 8072
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011072054516193317,
+      "loss": 0.7171,
+      "step": 8073
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011070245982859828,
+      "loss": 0.6686,
+      "step": 8074
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001106843741411451,
+      "loss": 0.7038,
+      "step": 8075
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011066628810017206,
+      "loss": 0.7117,
+      "step": 8076
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001106482017062775,
+      "loss": 0.7776,
+      "step": 8077
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011063011496005992,
+      "loss": 0.7477,
+      "step": 8078
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011061202786211773,
+      "loss": 0.7091,
+      "step": 8079
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001105939404130494,
+      "loss": 0.7028,
+      "step": 8080
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001105758526134534,
+      "loss": 0.7024,
+      "step": 8081
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011055776446392825,
+      "loss": 0.7235,
+      "step": 8082
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001105396759650724,
+      "loss": 0.6343,
+      "step": 8083
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011052158711748434,
+      "loss": 0.7077,
+      "step": 8084
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011050349792176263,
+      "loss": 0.6717,
+      "step": 8085
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011048540837850572,
+      "loss": 0.7239,
+      "step": 8086
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011046731848831226,
+      "loss": 0.6768,
+      "step": 8087
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011044922825178074,
+      "loss": 0.7287,
+      "step": 8088
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011043113766950972,
+      "loss": 0.7294,
+      "step": 8089
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011041304674209778,
+      "loss": 0.6859,
+      "step": 8090
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011039495547014349,
+      "loss": 0.691,
+      "step": 8091
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011037686385424548,
+      "loss": 0.6549,
+      "step": 8092
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011035877189500232,
+      "loss": 0.6459,
+      "step": 8093
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011034067959301266,
+      "loss": 0.6599,
+      "step": 8094
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001103225869488751,
+      "loss": 0.6773,
+      "step": 8095
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011030449396318831,
+      "loss": 0.6685,
+      "step": 8096
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001102864006365509,
+      "loss": 0.7301,
+      "step": 8097
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00011026830696956164,
+      "loss": 0.7016,
+      "step": 8098
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011025021296281907,
+      "loss": 0.7345,
+      "step": 8099
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011023211861692196,
+      "loss": 0.681,
+      "step": 8100
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011021402393246897,
+      "loss": 0.624,
+      "step": 8101
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011019592891005883,
+      "loss": 0.6248,
+      "step": 8102
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011017783355029026,
+      "loss": 0.6593,
+      "step": 8103
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011015973785376199,
+      "loss": 0.6586,
+      "step": 8104
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011014164182107274,
+      "loss": 0.6521,
+      "step": 8105
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011012354545282129,
+      "loss": 0.6284,
+      "step": 8106
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0001101054487496064,
+      "loss": 0.7214,
+      "step": 8107
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011008735171202684,
+      "loss": 0.7505,
+      "step": 8108
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011006925434068139,
+      "loss": 0.7861,
+      "step": 8109
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011005115663616887,
+      "loss": 0.6824,
+      "step": 8110
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011003305859908806,
+      "loss": 0.7773,
+      "step": 8111
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00011001496023003781,
+      "loss": 0.6375,
+      "step": 8112
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010999686152961693,
+      "loss": 0.6496,
+      "step": 8113
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010997876249842427,
+      "loss": 0.659,
+      "step": 8114
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010996066313705869,
+      "loss": 0.7126,
+      "step": 8115
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010994256344611901,
+      "loss": 0.6956,
+      "step": 8116
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010992446342620419,
+      "loss": 0.6905,
+      "step": 8117
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010990636307791305,
+      "loss": 0.6988,
+      "step": 8118
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010988826240184446,
+      "loss": 0.7258,
+      "step": 8119
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010987016139859743,
+      "loss": 0.6691,
+      "step": 8120
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010985206006877078,
+      "loss": 0.7577,
+      "step": 8121
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010983395841296348,
+      "loss": 0.7424,
+      "step": 8122
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010981585643177448,
+      "loss": 0.7032,
+      "step": 8123
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0001097977541258027,
+      "loss": 0.6765,
+      "step": 8124
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010977965149564712,
+      "loss": 0.6232,
+      "step": 8125
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010976154854190672,
+      "loss": 0.6616,
+      "step": 8126
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010974344526518045,
+      "loss": 0.686,
+      "step": 8127
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010972534166606734,
+      "loss": 0.785,
+      "step": 8128
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010970723774516638,
+      "loss": 0.7319,
+      "step": 8129
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0001096891335030766,
+      "loss": 0.6912,
+      "step": 8130
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010967102894039698,
+      "loss": 0.6434,
+      "step": 8131
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010965292405772658,
+      "loss": 0.7041,
+      "step": 8132
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010963481885566447,
+      "loss": 0.6893,
+      "step": 8133
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0001096167133348097,
+      "loss": 0.6877,
+      "step": 8134
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010959860749576131,
+      "loss": 0.6306,
+      "step": 8135
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010958050133911836,
+      "loss": 0.6877,
+      "step": 8136
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010956239486548003,
+      "loss": 0.6773,
+      "step": 8137
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010954428807544533,
+      "loss": 0.6534,
+      "step": 8138
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010952618096961341,
+      "loss": 0.6954,
+      "step": 8139
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010950807354858338,
+      "loss": 0.6242,
+      "step": 8140
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010948996581295436,
+      "loss": 0.702,
+      "step": 8141
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00010947185776332552,
+      "loss": 0.7417,
+      "step": 8142
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010945374940029598,
+      "loss": 0.6417,
+      "step": 8143
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010943564072446492,
+      "loss": 0.6999,
+      "step": 8144
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010941753173643152,
+      "loss": 0.6149,
+      "step": 8145
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010939942243679495,
+      "loss": 0.6996,
+      "step": 8146
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010938131282615436,
+      "loss": 0.6366,
+      "step": 8147
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010936320290510905,
+      "loss": 0.691,
+      "step": 8148
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010934509267425814,
+      "loss": 0.6847,
+      "step": 8149
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0001093269821342009,
+      "loss": 0.7271,
+      "step": 8150
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010930887128553658,
+      "loss": 0.6538,
+      "step": 8151
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010929076012886437,
+      "loss": 0.6708,
+      "step": 8152
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010927264866478353,
+      "loss": 0.7553,
+      "step": 8153
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010925453689389338,
+      "loss": 0.675,
+      "step": 8154
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010923642481679315,
+      "loss": 0.7143,
+      "step": 8155
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010921831243408218,
+      "loss": 0.7266,
+      "step": 8156
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010920019974635967,
+      "loss": 0.702,
+      "step": 8157
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010918208675422496,
+      "loss": 0.6959,
+      "step": 8158
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010916397345827742,
+      "loss": 0.6855,
+      "step": 8159
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010914585985911632,
+      "loss": 0.6964,
+      "step": 8160
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010912774595734101,
+      "loss": 0.6542,
+      "step": 8161
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010910963175355087,
+      "loss": 0.745,
+      "step": 8162
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010909151724834519,
+      "loss": 0.6694,
+      "step": 8163
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010907340244232331,
+      "loss": 0.7813,
+      "step": 8164
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010905528733608473,
+      "loss": 0.6923,
+      "step": 8165
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010903717193022873,
+      "loss": 0.637,
+      "step": 8166
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010901905622535477,
+      "loss": 0.6631,
+      "step": 8167
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0001090009402220622,
+      "loss": 0.6704,
+      "step": 8168
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010898282392095048,
+      "loss": 0.6899,
+      "step": 8169
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010896470732261897,
+      "loss": 0.7091,
+      "step": 8170
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010894659042766715,
+      "loss": 0.7052,
+      "step": 8171
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010892847323669451,
+      "loss": 0.6683,
+      "step": 8172
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010891035575030039,
+      "loss": 0.6978,
+      "step": 8173
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010889223796908437,
+      "loss": 0.7047,
+      "step": 8174
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010887411989364585,
+      "loss": 0.7094,
+      "step": 8175
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010885600152458428,
+      "loss": 0.7644,
+      "step": 8176
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0001088378828624993,
+      "loss": 0.6659,
+      "step": 8177
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010881976390799025,
+      "loss": 0.7085,
+      "step": 8178
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010880164466165674,
+      "loss": 0.6425,
+      "step": 8179
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010878352512409824,
+      "loss": 0.6973,
+      "step": 8180
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010876540529591431,
+      "loss": 0.647,
+      "step": 8181
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010874728517770447,
+      "loss": 0.6848,
+      "step": 8182
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010872916477006834,
+      "loss": 0.6609,
+      "step": 8183
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010871104407360537,
+      "loss": 0.6871,
+      "step": 8184
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00010869292308891521,
+      "loss": 0.7203,
+      "step": 8185
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001086748018165974,
+      "loss": 0.7136,
+      "step": 8186
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010865668025725151,
+      "loss": 0.711,
+      "step": 8187
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010863855841147723,
+      "loss": 0.6813,
+      "step": 8188
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010862043627987408,
+      "loss": 0.7102,
+      "step": 8189
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010860231386304172,
+      "loss": 0.7177,
+      "step": 8190
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010858419116157975,
+      "loss": 0.6764,
+      "step": 8191
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010856606817608781,
+      "loss": 0.6129,
+      "step": 8192
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010854794490716555,
+      "loss": 0.6558,
+      "step": 8193
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010852982135541266,
+      "loss": 0.672,
+      "step": 8194
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010851169752142874,
+      "loss": 0.6675,
+      "step": 8195
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010849357340581349,
+      "loss": 0.6947,
+      "step": 8196
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010847544900916665,
+      "loss": 0.7192,
+      "step": 8197
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010845732433208779,
+      "loss": 0.6242,
+      "step": 8198
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010843919937517672,
+      "loss": 0.6871,
+      "step": 8199
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010842107413903313,
+      "loss": 0.7309,
+      "step": 8200
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010840294862425669,
+      "loss": 0.6585,
+      "step": 8201
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010838482283144718,
+      "loss": 0.7555,
+      "step": 8202
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001083666967612043,
+      "loss": 0.6997,
+      "step": 8203
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010834857041412782,
+      "loss": 0.6268,
+      "step": 8204
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010833044379081752,
+      "loss": 0.7279,
+      "step": 8205
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010831231689187311,
+      "loss": 0.7541,
+      "step": 8206
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010829418971789442,
+      "loss": 0.6547,
+      "step": 8207
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010827606226948119,
+      "loss": 0.6713,
+      "step": 8208
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010825793454723325,
+      "loss": 0.6687,
+      "step": 8209
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010823980655175037,
+      "loss": 0.6374,
+      "step": 8210
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010822167828363236,
+      "loss": 0.7136,
+      "step": 8211
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010820354974347909,
+      "loss": 0.6852,
+      "step": 8212
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010818542093189032,
+      "loss": 0.6554,
+      "step": 8213
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010816729184946595,
+      "loss": 0.6693,
+      "step": 8214
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010814916249680578,
+      "loss": 0.7064,
+      "step": 8215
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001081310328745097,
+      "loss": 0.7005,
+      "step": 8216
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010811290298317755,
+      "loss": 0.5933,
+      "step": 8217
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010809477282340923,
+      "loss": 0.686,
+      "step": 8218
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001080766423958046,
+      "loss": 0.6579,
+      "step": 8219
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010805851170096354,
+      "loss": 0.7669,
+      "step": 8220
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010804038073948599,
+      "loss": 0.6395,
+      "step": 8221
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010802224951197184,
+      "loss": 0.6478,
+      "step": 8222
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010800411801902102,
+      "loss": 0.6647,
+      "step": 8223
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010798598626123341,
+      "loss": 0.6384,
+      "step": 8224
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010796785423920901,
+      "loss": 0.7805,
+      "step": 8225
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001079497219535477,
+      "loss": 0.6789,
+      "step": 8226
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001079315894048495,
+      "loss": 0.6212,
+      "step": 8227
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00010791345659371433,
+      "loss": 0.7083,
+      "step": 8228
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010789532352074216,
+      "loss": 0.6734,
+      "step": 8229
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.000107877190186533,
+      "loss": 0.6394,
+      "step": 8230
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010785905659168678,
+      "loss": 0.6886,
+      "step": 8231
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010784092273680358,
+      "loss": 0.6542,
+      "step": 8232
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010782278862248331,
+      "loss": 0.701,
+      "step": 8233
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010780465424932606,
+      "loss": 0.6623,
+      "step": 8234
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010778651961793183,
+      "loss": 0.7139,
+      "step": 8235
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010776838472890065,
+      "loss": 0.7043,
+      "step": 8236
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010775024958283251,
+      "loss": 0.6197,
+      "step": 8237
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010773211418032754,
+      "loss": 0.6527,
+      "step": 8238
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010771397852198575,
+      "loss": 0.7472,
+      "step": 8239
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010769584260840722,
+      "loss": 0.722,
+      "step": 8240
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010767770644019202,
+      "loss": 0.6963,
+      "step": 8241
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010765957001794018,
+      "loss": 0.5639,
+      "step": 8242
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010764143334225187,
+      "loss": 0.7328,
+      "step": 8243
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010762329641372716,
+      "loss": 0.6689,
+      "step": 8244
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010760515923296617,
+      "loss": 0.6415,
+      "step": 8245
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010758702180056897,
+      "loss": 0.7079,
+      "step": 8246
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010756888411713575,
+      "loss": 0.6447,
+      "step": 8247
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010755074618326655,
+      "loss": 0.706,
+      "step": 8248
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010753260799956157,
+      "loss": 0.6274,
+      "step": 8249
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010751446956662097,
+      "loss": 0.6744,
+      "step": 8250
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0001074963308850449,
+      "loss": 0.6736,
+      "step": 8251
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010747819195543351,
+      "loss": 0.7157,
+      "step": 8252
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010746005277838697,
+      "loss": 0.6794,
+      "step": 8253
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010744191335450545,
+      "loss": 0.6904,
+      "step": 8254
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010742377368438914,
+      "loss": 0.7314,
+      "step": 8255
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010740563376863828,
+      "loss": 0.6681,
+      "step": 8256
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010738749360785307,
+      "loss": 0.7281,
+      "step": 8257
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010736935320263371,
+      "loss": 0.7127,
+      "step": 8258
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010735121255358036,
+      "loss": 0.7043,
+      "step": 8259
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010733307166129331,
+      "loss": 0.6628,
+      "step": 8260
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010731493052637283,
+      "loss": 0.6817,
+      "step": 8261
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0001072967891494191,
+      "loss": 0.6616,
+      "step": 8262
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010727864753103245,
+      "loss": 0.7465,
+      "step": 8263
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010726050567181304,
+      "loss": 0.7186,
+      "step": 8264
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010724236357236121,
+      "loss": 0.6351,
+      "step": 8265
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010722422123327722,
+      "loss": 0.6789,
+      "step": 8266
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010720607865516137,
+      "loss": 0.8318,
+      "step": 8267
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010718793583861395,
+      "loss": 0.6163,
+      "step": 8268
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010716979278423522,
+      "loss": 0.688,
+      "step": 8269
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010715164949262555,
+      "loss": 0.7363,
+      "step": 8270
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00010713350596438521,
+      "loss": 0.7195,
+      "step": 8271
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010711536220011455,
+      "loss": 0.7268,
+      "step": 8272
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010709721820041392,
+      "loss": 0.7369,
+      "step": 8273
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010707907396588361,
+      "loss": 0.7047,
+      "step": 8274
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010706092949712403,
+      "loss": 0.6907,
+      "step": 8275
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010704278479473546,
+      "loss": 0.684,
+      "step": 8276
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010702463985931832,
+      "loss": 0.657,
+      "step": 8277
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010700649469147299,
+      "loss": 0.6521,
+      "step": 8278
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010698834929179982,
+      "loss": 0.6919,
+      "step": 8279
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010697020366089921,
+      "loss": 0.6413,
+      "step": 8280
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010695205779937155,
+      "loss": 0.7237,
+      "step": 8281
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0001069339117078172,
+      "loss": 0.7021,
+      "step": 8282
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010691576538683665,
+      "loss": 0.7535,
+      "step": 8283
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010689761883703029,
+      "loss": 0.8037,
+      "step": 8284
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010687947205899853,
+      "loss": 0.6779,
+      "step": 8285
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010686132505334179,
+      "loss": 0.6649,
+      "step": 8286
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010684317782066053,
+      "loss": 0.7504,
+      "step": 8287
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010682503036155518,
+      "loss": 0.7088,
+      "step": 8288
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010680688267662627,
+      "loss": 0.6331,
+      "step": 8289
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010678873476647415,
+      "loss": 0.7262,
+      "step": 8290
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010677058663169936,
+      "loss": 0.6043,
+      "step": 8291
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010675243827290238,
+      "loss": 0.6974,
+      "step": 8292
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010673428969068364,
+      "loss": 0.7201,
+      "step": 8293
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0001067161408856437,
+      "loss": 0.7872,
+      "step": 8294
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010669799185838301,
+      "loss": 0.6619,
+      "step": 8295
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0001066798426095021,
+      "loss": 0.6456,
+      "step": 8296
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010666169313960148,
+      "loss": 0.6769,
+      "step": 8297
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010664354344928167,
+      "loss": 0.6285,
+      "step": 8298
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010662539353914316,
+      "loss": 0.6802,
+      "step": 8299
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010660724340978661,
+      "loss": 0.6346,
+      "step": 8300
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010658909306181241,
+      "loss": 0.6704,
+      "step": 8301
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010657094249582122,
+      "loss": 0.6677,
+      "step": 8302
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010655279171241354,
+      "loss": 0.7321,
+      "step": 8303
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010653464071218998,
+      "loss": 0.7376,
+      "step": 8304
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010651648949575106,
+      "loss": 0.6986,
+      "step": 8305
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010649833806369738,
+      "loss": 0.7201,
+      "step": 8306
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010648018641662955,
+      "loss": 0.6663,
+      "step": 8307
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010646203455514814,
+      "loss": 0.7277,
+      "step": 8308
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010644388247985377,
+      "loss": 0.697,
+      "step": 8309
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010642573019134703,
+      "loss": 0.7155,
+      "step": 8310
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010640757769022854,
+      "loss": 0.7266,
+      "step": 8311
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0001063894249770989,
+      "loss": 0.7297,
+      "step": 8312
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010637127205255877,
+      "loss": 0.7126,
+      "step": 8313
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00010635311891720878,
+      "loss": 0.6257,
+      "step": 8314
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010633496557164959,
+      "loss": 0.6759,
+      "step": 8315
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010631681201648181,
+      "loss": 0.6845,
+      "step": 8316
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010629865825230611,
+      "loss": 0.7502,
+      "step": 8317
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010628050427972318,
+      "loss": 0.6321,
+      "step": 8318
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010626235009933367,
+      "loss": 0.6941,
+      "step": 8319
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010624419571173825,
+      "loss": 0.6794,
+      "step": 8320
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010622604111753762,
+      "loss": 0.6754,
+      "step": 8321
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010620788631733246,
+      "loss": 0.6853,
+      "step": 8322
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010618973131172347,
+      "loss": 0.6918,
+      "step": 8323
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010617157610131136,
+      "loss": 0.6938,
+      "step": 8324
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010615342068669685,
+      "loss": 0.6832,
+      "step": 8325
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010613526506848063,
+      "loss": 0.6364,
+      "step": 8326
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010611710924726346,
+      "loss": 0.7301,
+      "step": 8327
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010609895322364605,
+      "loss": 0.7222,
+      "step": 8328
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010608079699822915,
+      "loss": 0.682,
+      "step": 8329
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010606264057161348,
+      "loss": 0.6116,
+      "step": 8330
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010604448394439983,
+      "loss": 0.6881,
+      "step": 8331
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010602632711718892,
+      "loss": 0.7192,
+      "step": 8332
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010600817009058154,
+      "loss": 0.6394,
+      "step": 8333
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010599001286517847,
+      "loss": 0.691,
+      "step": 8334
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010597185544158047,
+      "loss": 0.7509,
+      "step": 8335
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010595369782038831,
+      "loss": 0.6598,
+      "step": 8336
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010593554000220285,
+      "loss": 0.6843,
+      "step": 8337
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010591738198762478,
+      "loss": 0.6958,
+      "step": 8338
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010589922377725498,
+      "loss": 0.6656,
+      "step": 8339
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010588106537169425,
+      "loss": 0.7149,
+      "step": 8340
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001058629067715434,
+      "loss": 0.6553,
+      "step": 8341
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001058447479774033,
+      "loss": 0.7245,
+      "step": 8342
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010582658898987468,
+      "loss": 0.7101,
+      "step": 8343
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001058084298095584,
+      "loss": 0.6668,
+      "step": 8344
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001057902704370554,
+      "loss": 0.7407,
+      "step": 8345
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010577211087296644,
+      "loss": 0.6894,
+      "step": 8346
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010575395111789241,
+      "loss": 0.6628,
+      "step": 8347
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010573579117243417,
+      "loss": 0.6668,
+      "step": 8348
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010571763103719254,
+      "loss": 0.6541,
+      "step": 8349
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010569947071276847,
+      "loss": 0.7075,
+      "step": 8350
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001056813101997628,
+      "loss": 0.6501,
+      "step": 8351
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010566314949877643,
+      "loss": 0.7176,
+      "step": 8352
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010564498861041026,
+      "loss": 0.7601,
+      "step": 8353
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010562682753526513,
+      "loss": 0.6389,
+      "step": 8354
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00010560866627394202,
+      "loss": 0.6275,
+      "step": 8355
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001055905048270418,
+      "loss": 0.7173,
+      "step": 8356
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001055723431951654,
+      "loss": 0.6776,
+      "step": 8357
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001055541813789138,
+      "loss": 0.6913,
+      "step": 8358
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010553601937888782,
+      "loss": 0.7603,
+      "step": 8359
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010551785719568848,
+      "loss": 0.6656,
+      "step": 8360
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010549969482991664,
+      "loss": 0.6786,
+      "step": 8361
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010548153228217335,
+      "loss": 0.6842,
+      "step": 8362
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010546336955305952,
+      "loss": 0.7066,
+      "step": 8363
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010544520664317613,
+      "loss": 0.6807,
+      "step": 8364
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010542704355312407,
+      "loss": 0.651,
+      "step": 8365
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010540888028350438,
+      "loss": 0.6901,
+      "step": 8366
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010539071683491804,
+      "loss": 0.6534,
+      "step": 8367
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.000105372553207966,
+      "loss": 0.6788,
+      "step": 8368
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0001053543894032493,
+      "loss": 0.7333,
+      "step": 8369
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010533622542136889,
+      "loss": 0.7008,
+      "step": 8370
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0001053180612629258,
+      "loss": 0.6887,
+      "step": 8371
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010529989692852098,
+      "loss": 0.686,
+      "step": 8372
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010528173241875552,
+      "loss": 0.7134,
+      "step": 8373
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010526356773423043,
+      "loss": 0.7819,
+      "step": 8374
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0001052454028755467,
+      "loss": 0.7253,
+      "step": 8375
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010522723784330537,
+      "loss": 0.6562,
+      "step": 8376
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010520907263810746,
+      "loss": 0.6793,
+      "step": 8377
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010519090726055402,
+      "loss": 0.6196,
+      "step": 8378
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010517274171124616,
+      "loss": 0.6652,
+      "step": 8379
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010515457599078488,
+      "loss": 0.5998,
+      "step": 8380
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010513641009977124,
+      "loss": 0.7147,
+      "step": 8381
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010511824403880629,
+      "loss": 0.6939,
+      "step": 8382
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010510007780849112,
+      "loss": 0.6818,
+      "step": 8383
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0001050819114094268,
+      "loss": 0.6993,
+      "step": 8384
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010506374484221446,
+      "loss": 0.5839,
+      "step": 8385
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010504557810745512,
+      "loss": 0.7149,
+      "step": 8386
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010502741120574991,
+      "loss": 0.6712,
+      "step": 8387
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010500924413769988,
+      "loss": 0.6758,
+      "step": 8388
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010499107690390617,
+      "loss": 0.7378,
+      "step": 8389
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010497290950496994,
+      "loss": 0.6985,
+      "step": 8390
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010495474194149224,
+      "loss": 0.6228,
+      "step": 8391
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010493657421407418,
+      "loss": 0.7124,
+      "step": 8392
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010491840632331695,
+      "loss": 0.6743,
+      "step": 8393
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010490023826982163,
+      "loss": 0.6738,
+      "step": 8394
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010488207005418935,
+      "loss": 0.6872,
+      "step": 8395
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0001048639016770213,
+      "loss": 0.7065,
+      "step": 8396
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010484573313891857,
+      "loss": 0.7599,
+      "step": 8397
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010482756444048237,
+      "loss": 0.6753,
+      "step": 8398
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010480939558231379,
+      "loss": 0.6723,
+      "step": 8399
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010479122656501406,
+      "loss": 0.6823,
+      "step": 8400
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00010477305738918433,
+      "loss": 0.6904,
+      "step": 8401
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010475488805542575,
+      "loss": 0.6758,
+      "step": 8402
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010473671856433951,
+      "loss": 0.6922,
+      "step": 8403
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0001047185489165268,
+      "loss": 0.6571,
+      "step": 8404
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010470037911258881,
+      "loss": 0.6249,
+      "step": 8405
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010468220915312671,
+      "loss": 0.6921,
+      "step": 8406
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010466403903874176,
+      "loss": 0.6832,
+      "step": 8407
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010464586877003513,
+      "loss": 0.656,
+      "step": 8408
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.000104627698347608,
+      "loss": 0.6789,
+      "step": 8409
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010460952777206161,
+      "loss": 0.7121,
+      "step": 8410
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010459135704399718,
+      "loss": 0.7803,
+      "step": 8411
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010457318616401597,
+      "loss": 0.6777,
+      "step": 8412
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010455501513271915,
+      "loss": 0.6496,
+      "step": 8413
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010453684395070797,
+      "loss": 0.7006,
+      "step": 8414
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0001045186726185837,
+      "loss": 0.7121,
+      "step": 8415
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010450050113694756,
+      "loss": 0.6528,
+      "step": 8416
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010448232950640082,
+      "loss": 0.6567,
+      "step": 8417
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0001044641577275447,
+      "loss": 0.7506,
+      "step": 8418
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0001044459858009805,
+      "loss": 0.681,
+      "step": 8419
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010442781372730946,
+      "loss": 0.715,
+      "step": 8420
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010440964150713287,
+      "loss": 0.6705,
+      "step": 8421
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010439146914105196,
+      "loss": 0.6896,
+      "step": 8422
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010437329662966807,
+      "loss": 0.7321,
+      "step": 8423
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010435512397358243,
+      "loss": 0.7287,
+      "step": 8424
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010433695117339637,
+      "loss": 0.7273,
+      "step": 8425
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010431877822971117,
+      "loss": 0.6971,
+      "step": 8426
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010430060514312811,
+      "loss": 0.6806,
+      "step": 8427
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010428243191424851,
+      "loss": 0.7221,
+      "step": 8428
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010426425854367367,
+      "loss": 0.7259,
+      "step": 8429
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010424608503200492,
+      "loss": 0.6506,
+      "step": 8430
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010422791137984359,
+      "loss": 0.7101,
+      "step": 8431
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010420973758779094,
+      "loss": 0.7299,
+      "step": 8432
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010419156365644833,
+      "loss": 0.7285,
+      "step": 8433
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010417338958641712,
+      "loss": 0.6947,
+      "step": 8434
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010415521537829859,
+      "loss": 0.6408,
+      "step": 8435
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010413704103269415,
+      "loss": 0.6731,
+      "step": 8436
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010411886655020508,
+      "loss": 0.6463,
+      "step": 8437
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010410069193143279,
+      "loss": 0.7064,
+      "step": 8438
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010408251717697857,
+      "loss": 0.7149,
+      "step": 8439
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010406434228744379,
+      "loss": 0.6816,
+      "step": 8440
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010404616726342986,
+      "loss": 0.6858,
+      "step": 8441
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0001040279921055381,
+      "loss": 0.7534,
+      "step": 8442
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010400981681436994,
+      "loss": 0.6977,
+      "step": 8443
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00010399164139052668,
+      "loss": 0.5958,
+      "step": 8444
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010397346583460971,
+      "loss": 0.643,
+      "step": 8445
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010395529014722048,
+      "loss": 0.6801,
+      "step": 8446
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010393711432896031,
+      "loss": 0.7613,
+      "step": 8447
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010391893838043065,
+      "loss": 0.6704,
+      "step": 8448
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010390076230223288,
+      "loss": 0.7447,
+      "step": 8449
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010388258609496835,
+      "loss": 0.6921,
+      "step": 8450
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010386440975923855,
+      "loss": 0.6715,
+      "step": 8451
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010384623329564483,
+      "loss": 0.7416,
+      "step": 8452
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010382805670478863,
+      "loss": 0.6965,
+      "step": 8453
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010380987998727139,
+      "loss": 0.7133,
+      "step": 8454
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010379170314369448,
+      "loss": 0.6827,
+      "step": 8455
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010377352617465934,
+      "loss": 0.6551,
+      "step": 8456
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010375534908076742,
+      "loss": 0.7657,
+      "step": 8457
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010373717186262018,
+      "loss": 0.6591,
+      "step": 8458
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010371899452081904,
+      "loss": 0.6393,
+      "step": 8459
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010370081705596542,
+      "loss": 0.6655,
+      "step": 8460
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010368263946866076,
+      "loss": 0.65,
+      "step": 8461
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010366446175950653,
+      "loss": 0.6843,
+      "step": 8462
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010364628392910423,
+      "loss": 0.7438,
+      "step": 8463
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010362810597805526,
+      "loss": 0.6773,
+      "step": 8464
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010360992790696114,
+      "loss": 0.6992,
+      "step": 8465
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010359174971642327,
+      "loss": 0.6883,
+      "step": 8466
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010357357140704316,
+      "loss": 0.7006,
+      "step": 8467
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010355539297942227,
+      "loss": 0.6937,
+      "step": 8468
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010353721443416211,
+      "loss": 0.6548,
+      "step": 8469
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010351903577186417,
+      "loss": 0.6632,
+      "step": 8470
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0001035008569931299,
+      "loss": 0.7242,
+      "step": 8471
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0001034826780985608,
+      "loss": 0.6884,
+      "step": 8472
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010346449908875833,
+      "loss": 0.7277,
+      "step": 8473
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010344631996432408,
+      "loss": 0.6712,
+      "step": 8474
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010342814072585949,
+      "loss": 0.6608,
+      "step": 8475
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010340996137396608,
+      "loss": 0.7273,
+      "step": 8476
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00010339178190924537,
+      "loss": 0.7362,
+      "step": 8477
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 17276,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 4,
+  "save_steps": 173,
+  "total_flos": 1.913647968842416e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}